Repository Analysis

rllm-org/rllm

Democratizing Reinforcement Learning for LLMs

50.3 Strong AI signal View on GitHub

Analysis Overview

This report presents the forensic synthetic code analysis of rllm-org/rllm, a Python project with 5,694 GitHub stars. SynthScan v2.0 examined 108,416 lines of code across 646 source files, recording 3084 pattern matches distributed across 23 syntactic categories. The overall adjusted score of 50.3 places this repository in the Strong AI signal band.

The scanner applied 160+ deterministic lexical heuristics, multi-line block detectors, abstract syntax tree depth profilers, and a cross-file Jaccard similarity matrix to construct a statistically normalised synthetic code estimate. All matches are individually weighted by severity coefficient and contextual multiplier before summation, and the resulting headline score is temporally discounted to account for the repository's development history relative to the commercial emergence of large language model coding tooling (November 2022 onward).

50.3
Adjusted Score
50.3
Raw Score
100%
Time Factor
2026-07-14
Last Push
5.7K
Stars
Python
Language
108.4K
Lines of Code
646
Files
3.1K
Pattern Hits
2026-07-14
Scan Date
0.14
HC Hit Rate

What These Metrics Mean

Adjusted Score
Primary synthetic code indicator. Raw score normalised per 1,000 lines of code and multiplied by the temporal discount factor. This is the definitive comparative metric — use it to rank repositories by AI authorship density.
Raw Score
The unmodified sum of all severity-weighted, context-multiplied pattern match scores before temporal discounting. Reflects the absolute signal strength independent of when the repository was last active.
Time Factor
The temporal discount multiplier (0–100%) applied to the raw score. Repositories last updated before ChatGPT's launch (Nov 2022) receive a 5% factor. Full signal is only assigned to repositories active in the post-adoption era (Jan 2024+).
Pattern Hits
Total count of individual pattern matches across all files and categories. A high hit count with a low score may indicate a very large codebase with isolated AI snippets; a low count with a high score indicates dense, concentrated AI signatures.
HC Hit Rate
High+Critical pattern hits per file, averaged across the repository. This orthogonal signal catches repositories where a few files are densely packed with high-severity AI tells — a strong indicator even when the normalised score appears moderate due to codebase size.
Lines of Code / Files
Total lines and files analysed. The scanner examines 94 file extensions. These denominators are used to normalise the score, enabling fair comparison between repositories of vastly different sizes.

Score History

This chart maps the temporal evolution of the adjusted synthetic code score across successive scan runs. An upward trajectory indicates ongoing incorporation of AI-generated code or expanding LLM-assisted scaffolding; a stable or declining trajectory may reflect active human refactoring, code removal, or the adoption of stricter authorship policies. The dashed secondary line (right axis) independently tracks total raw pattern hit count, which can diverge from the normalised score when codebase size changes significantly between scans.

Severity Breakdown

Classifies detected patterns by their diagnostic confidence and structural impact. CRITICAL patterns (coefficient 10) represent definitive synthetic signatures — hallucinated imports, explicit LLM attribution metadata — virtually never produced by human authors. HIGH (5) indicates strong structural tells such as cross-file repetition or cross-linguistic idioms. MEDIUM (2) covers recognisable conversational padding and AI-specific vocabulary. LOW (1) captures subtle indicators like tautological comments and generic boilerplate that require density to carry independent signal.

CRITICAL 0HIGH 89MEDIUM 770LOW 2225

Directory Score Breakdown

This horizontal bar chart decomposes the repository's raw synthetic code score by top-level directory, allowing you to pinpoint precisely which modules or components carry the highest AI authorship density. Directories with disproportionately high scores relative to their size warrant targeted manual review: concentrated AI signatures often trace back to mass-generated configuration layers, auto-ported test suites, LLM-scaffolded boilerplate classes, or entire subsystems authored under heavy copilot assistance. Use this view to prioritise your human code-review effort.

Pattern Findings

The scanner identified 3084 distinct pattern matches across 23 syntactic categories. Each entry below represents a discrete location in the source code where the engine recorded a statistically significant AI authorship indicator. Expand any category row to inspect the individual file paths, line numbers, code snippets, and the lexical context (CODE, COMMENT, or STRING) in which each match was detected.

Reading the findings table: The Severity column indicates the diagnostic confidence level (CRITICAL / HIGH / MEDIUM / LOW). The Context column identifies whether the match occurred inside executable code, an inline comment, or a string literal — comment-context matches receive a ×1.5 weight because LLMs systematically over-annotate. The ⚡ bolt icon marks clustered matches: three or more patterns within a 10-line window, each receiving an additional ×1.5 density multiplier as dense clusters constitute far stronger evidence of synthetic authorship than isolated hits.

Decorative Section Separators662 hits · 2314 pts
SeverityFileLineSnippetContext
MEDIUMtests/test_store.py11# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_store.py13# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_store.py71# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_store.py73# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_store.py116# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_store.py118# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_store.py168# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_store.py170# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_verl_policy_loss.py11# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_verl_policy_loss.py13# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_verl_policy_loss.py73# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_verl_policy_loss.py75# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_verl_policy_loss.py95# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_verl_policy_loss.py97# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_verl_policy_loss.py138# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_verl_policy_loss.py140# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_verl_policy_loss.py160# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_verl_policy_loss.py162# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_verl_policy_loss.py202# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_verl_policy_loss.py204# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_verl_policy_loss.py228# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/test_verl_policy_loss.py230# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/integration/test_remote_engine.py23# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/integration/test_remote_engine.py25# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/integration/test_remote_engine.py66# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/integration/test_remote_engine.py68# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/integration/test_sampling_e2e.py24# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/integration/test_sampling_e2e.py26# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/integration/test_sampling_e2e.py98# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/integration/test_sampling_e2e.py101# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/integration/test_sampling_e2e.py130# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/integration/test_sampling_e2e.py132# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/integration/test_sampling_e2e.py158# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/integration/test_sampling_e2e.py160# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/unified_trainer/test_tinker_transform.py145# =============================================================================COMMENT
MEDIUMtests/unified_trainer/test_tinker_transform.py147# =============================================================================COMMENT
MEDIUMtests/unified_trainer/test_tinker_transform.py199# =============================================================================COMMENT
MEDIUMtests/unified_trainer/test_tinker_transform.py201# =============================================================================COMMENT
MEDIUMtests/unified_trainer/test_tinker_transform.py267# =============================================================================COMMENT
MEDIUMtests/unified_trainer/test_tinker_transform.py269# =============================================================================COMMENT
MEDIUMtests/unified_trainer/test_tinker_transform.py323# =============================================================================COMMENT
MEDIUMtests/unified_trainer/test_tinker_transform.py325# =============================================================================COMMENT
MEDIUMtests/unified_trainer/test_tinker_transform.py361# =============================================================================COMMENT
MEDIUMtests/unified_trainer/test_tinker_transform.py363# =============================================================================COMMENT
MEDIUMtests/unified_trainer/test_tinker_transform.py21# =============================================================================COMMENT
MEDIUMtests/unified_trainer/test_tinker_transform.py23# =============================================================================COMMENT
MEDIUMtests/unified_trainer/test_tinker_transform.py78# =============================================================================COMMENT
MEDIUMtests/unified_trainer/test_tinker_transform.py80# =============================================================================COMMENT
MEDIUMtests/unified_trainer/test_tinker_transform.py464# =============================================================================COMMENT
MEDIUMtests/unified_trainer/test_tinker_transform.py466# =============================================================================COMMENT
MEDIUMtests/unified_trainer/test_config_sync.py105# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/unified_trainer/test_config_sync.py107# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/unified_trainer/test_config_sync.py196# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/unified_trainer/test_config_sync.py198# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/unified_trainer/test_config_sync.py262# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/unified_trainer/test_config_sync.py264# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/unified_trainer/test_config_sync.py302# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/unified_trainer/test_config_sync.py304# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/harnesses/test_cli_harness.py79# ---------------------------------------------------------------------------COMMENT
MEDIUMtests/harnesses/test_cli_harness.py82# ---------------------------------------------------------------------------COMMENT
602 more matches not shown…
Hyper-Verbose Identifiers1011 hits · 1100 pts
SeverityFileLineSnippetContext
LOWtests/test_logging.py28def test_get_log_level_from_env_uses_default(monkeypatch):CODE
LOWtests/test_logging.py35def test_get_log_level_from_env_reads_level(monkeypatch):CODE
LOWtests/test_logging.py41def test_get_log_level_from_env_falls_back_for_invalid_level(monkeypatch):CODE
LOWtests/test_logging.py47def test_configure_logging_from_env_sets_rllm_logger(monkeypatch):CODE
LOWtests/test_logging.py55def test_configure_logging_from_env_initializes_basic_logging(monkeypatch):CODE
LOWtests/test_logging.py68def test_import_configures_rllm_logger_from_env(monkeypatch):CODE
LOWtests/test_verl_imports.py47def test_verl_module_paths_exist(module_path: str, name: str) -> None:CODE
LOWtests/test_verl_metrics.py24def test_calculate_debug_metrics_compat_passes_through_upstream_metrics(monkeypatch: pytest.MonkeyPatch) -> None:CODE
LOWtests/test_verl_metrics.py49def test_calculate_debug_metrics_compat_keeps_newer_upstream_empty_mask_fallback(CODE
LOWtests/test_verl_metrics.py72def test_calculate_debug_metrics_compat_single_token_std_is_zero(monkeypatch: pytest.MonkeyPatch) -> None:CODE
LOWtests/test_verl_metrics.py90def test_calculate_debug_metrics_compat_backfills_legacy_empty_mask_error(CODE
LOWtests/test_verl_metrics.py113def test_calculate_debug_metrics_compat_reraises_unrelated_runtime_errors(CODE
LOWtests/test_store.py27 def test_get_missing_key_returns_default(self, store):CODE
LOWtests/test_store.py64 def test_stores_arbitrary_types(self, store):CODE
LOWtests/test_store.py77 def test_concurrent_writes_no_corruption(self, store):CODE
LOWtests/test_store.py94 def test_concurrent_reads_during_writes(self, store):CODE
LOWtests/test_store.py122 def test_two_instances_share_state(self, store):CODE
LOWtests/test_store.py138 def test_interleaved_read_write(self, store):CODE
LOWtests/test_store.py174 def test_workflow_receives_store(self):CODE
LOWtests/test_store.py192 def test_workflow_without_store(self):CODE
LOWtests/test_verl_policy_loss.py17 def test_bare_values_unchanged(self):CODE
LOWtests/test_verl_policy_loss.py79 def test_unknown_loss_triggers_warning(self, caplog):CODE
LOWtests/test_verl_policy_loss.py88 def test_known_losses_complete(self):CODE
LOWtests/test_verl_policy_loss.py108 def test_override_applied_and_restored(self):CODE
LOWtests/test_verl_policy_loss.py130 def test_no_override_passes_through(self):CODE
LOWtests/test_verl_policy_loss.py151 def test_default_loss_is_known(self):CODE
LOWtests/test_verl_policy_loss.py166 def test_roles_with_same_loss_grouped(self):CODE
LOWtests/test_verl_policy_loss.py208 def test_accumulated_data_tracks_group_roles(self):CODE
LOWtests/test_verl_policy_loss.py234 def test_no_loss_fn_map_means_no_splitting(self):CODE
LOWtests/test_verl_policy_loss.py53 def test_invalid_tuple_length(self):CODE
LOWtests/test_verl_policy_loss.py62 def test_backward_compat_no_tuples(self):CODE
LOWtests/test_verl_policy_loss.py179 def test_all_same_loss_single_group(self):CODE
LOWtests/test_verl_policy_loss.py190 def test_each_role_different_loss(self):CODE
LOWtests/test_verl_policy_loss.py245 def test_advantage_computation_unaffected(self):CODE
LOWtests/tasks/test_loader_memory.py8def test_vlm_instruction_encodes_images_lazily(tmp_path, monkeypatch):CODE
LOWtests/integration/test_remote_engine.py105 async def test_batch_tasks_session_correlation(self, engine):CODE
LOWtests/integration/test_remote_engine.py130 async def test_timeout_produces_error_episode(self, engine):CODE
LOWtests/integration/test_sampling_e2e.py135def test_eval_run_dataset_enforces_sampling(mock_upstream):CODE
LOWtests/integration/test_sampling_e2e.py163def test_train_rollout_enforces_train_sampling(mock_upstream):CODE
LOWtests/integration/test_minimax_integration.py27 def test_minimax_litellm_config_generation(self):CODE
LOWtests/integration/test_minimax_integration.py41 def test_minimax_config_save_and_load(self, tmp_path, monkeypatch):CODE
LOWtests/integration/test_minimax_integration.py60 def test_minimax_litellm_completion(self):CODE
LOWtests/integration/test_tinker_gateway_integration.py124 def test_vllm_fields_stripped(self, tinker_gateway):CODE
LOWtests/integration/test_tinker_gateway_integration.py147 def test_trace_has_token_data(self, tinker_gateway):CODE
LOWtests/integration/test_tinker_gateway_integration.py177 def test_trace_round_trips_to_step(self, tinker_gateway):CODE
LOWtests/integration/test_tinker_gateway_integration.py211 def test_streaming_delivers_content_and_trace(self, tinker_gateway):CODE
LOWtests/integration/test_tinker_gateway_integration.py265 def test_tool_calling_multi_turn(self, tinker_gateway):CODE
LOWtests/unified_trainer/test_warm_queue_wiring.py15def test_detached_warm_queue_detaches_and_restores():CODE
LOWtests/unified_trainer/test_warm_queue_wiring.py53def test_start_warm_queue_guard_returns_none(warm_queue_size, sandbox_backend, hooks_present):CODE
LOWtests/unified_trainer/test_algorithm_config.py48def test_norm_adv_by_std_in_grpo_true_from_algorithm():CODE
LOWtests/unified_trainer/test_algorithm_config.py55def test_norm_adv_by_std_in_grpo_false_from_algorithm():CODE
LOWtests/unified_trainer/test_algorithm_config.py62def test_warmup_steps_from_algorithm():CODE
LOWtests/unified_trainer/test_termination_metrics.py6def test_collect_workflow_metrics_counts_unknown_termination_reason():CODE
LOWtests/unified_trainer/test_tinker_checkpoint.py37def test_auto_no_checkpoint_returns_none(tmp_path):CODE
LOWtests/unified_trainer/test_tinker_checkpoint.py41def test_auto_resumes_latest_from_tracker(tmp_path):CODE
LOWtests/unified_trainer/test_tinker_checkpoint.py47def test_auto_tracker_points_at_missing_dir_returns_none(tmp_path):CODE
LOWtests/unified_trainer/test_tinker_checkpoint.py53def test_disable_returns_none_even_with_checkpoint(tmp_path):CODE
LOWtests/unified_trainer/test_tinker_checkpoint.py58def test_resume_path_explicit_folder(tmp_path):CODE
LOWtests/unified_trainer/test_tinker_checkpoint.py63def test_resume_path_missing_checkpoint_raises(tmp_path):CODE
LOWtests/unified_trainer/test_tinker_checkpoint.py69def test_resume_path_requires_global_step_token(tmp_path):CODE
951 more matches not shown…
Excessive Try-Catch Wrapping413 hits · 470 pts
SeverityFileLineSnippetContext
LOWtests/conftest.py22 except Exception:CODE
LOWtests/conftest.py31 except Exception:CODE
LOWtests/conftest.py51 except Exception:CODE
LOWtests/conftest.py121 except Exception:CODE
MEDIUMtests/tools/test_tools.py143 print(f"Error in tool execution: {result.error}")CODE
MEDIUMtests/tools/test_tools.py175 print(f"Error: {result.error}")CODE
MEDIUMtests/tools/test_tools.py204 print(f"Error: {result.error}")CODE
MEDIUMtests/tools/test_tools.py245 print(f"Error: {result.error}")CODE
LOWtests/parser/test_harmony_chat_parser.py17 except Exception as exc:CODE
LOWtests/parser/test_chat_parser.py111 except Exception as e:CODE
LOWrllm/hooks.py237 except Exception:CODE
LOWrllm/hooks.py267 except Exception as e:CODE
LOWrllm/hooks.py277 except Exception:CODE
LOWrllm/hooks.py288 except Exception:CODE
LOW…ental/rllm_telemetry/examples/experiment_comparison.py107 except Exception as e:CODE
LOW…erimental/rllm_telemetry/src/rllm_telemetry/scorers.py250 except Exception as exc:CODE
LOW…rllm_telemetry/src/rllm_telemetry/trajectory_export.py50 except Exception:CODE
LOW…rllm_telemetry/src/rllm_telemetry/trajectory_export.py55 except Exception:CODE
LOW…rllm_telemetry/src/rllm_telemetry/trajectory_export.py59 except Exception:CODE
MEDIUM…rimental/rllm_telemetry/src/rllm_telemetry/exporter.py425def _flush_loop(self) -> None:CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/exporter.py230 except Exception:CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/exporter.py378 except Exception as exc:CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/exporter.py385 except Exception as exc:CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/exporter.py394 except Exception:CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/exporter.py646 except Exception as exc:CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/exporter.py407 except Exception:CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/exporter.py457 except Exception as exc:CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/exporter.py474 except Exception:CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/exporter.py563 except Exception as exc:CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/exporter.py584 except Exception as exc:CODE
MEDIUM…perimental/rllm_telemetry/src/rllm_telemetry/plugin.py63def wrapper(self, **kwargs):CODE
LOW…perimental/rllm_telemetry/src/rllm_telemetry/plugin.py66 except Exception:CODE
LOW…perimental/rllm_telemetry/src/rllm_telemetry/plugin.py159 except Exception:CODE
LOW…perimental/rllm_telemetry/src/rllm_telemetry/plugin.py163 except Exception:CODE
LOW…perimental/rllm_telemetry/src/rllm_telemetry/plugin.py91 except Exception:CODE
LOW…experimental/rllm_telemetry/src/rllm_telemetry/eval.py285 except Exception as exc:CODE
LOW…experimental/rllm_telemetry/src/rllm_telemetry/eval.py300 except Exception as exc:CODE
LOWrllm/experimental/fully_async/runner.py246 except Exception as e:CODE
LOWrllm/experimental/fully_async/runner.py254 except Exception as e:CODE
LOWrllm/experimental/fully_async/rollout_executor.py283 except Exception as e:STRING
LOWrllm/experimental/fully_async/rollout_executor.py303 except Exception as e:STRING
LOWrllm/experimental/fully_async/rollout_executor.py310 except Exception:STRING
LOWrllm/experimental/fully_async/rollout_executor.py388 except Exception:CODE
LOWrllm/experimental/fully_async/utils.py209 except Exception:CODE
LOWrllm/tasks/loader.py210 except Exception:CODE
LOWrllm/tasks/loader.py385 except Exception:CODE
LOWrllm/tools/utils.py208 except Exception as e:CODE
MEDIUMrllm/tools/utils.py213 print(f"Error in test case {{i+1}}: {{str(e)}}")CODE
LOWrllm/tools/utils.py284 except Exception as e:CODE
LOWrllm/tools/mcp_tool.py37 except Exception as e:CODE
MEDIUMrllm/tools/mcp_tool.py20def async_forward(self, **kwargs) -> ToolOutput:CODE
LOWrllm/tools/tool_base.py122 except Exception as e:CODE
LOWrllm/tools/code_tools/e2b_tool.py39 except Exception as e:CODE
MEDIUMrllm/tools/code_tools/e2b_tool.py40 print(f"Error killing sandbox: {e}")CODE
LOWrllm/tools/code_tools/e2b_tool.py77 except Exception:CODE
LOWrllm/tools/code_tools/together_tool.py90 except Exception as e:CODE
LOWrllm/tools/code_tools/lcb_tool.py52 except Exception as e:CODE
LOWrllm/tools/code_tools/lcb_tool.py99 except Exception as e:CODE
LOWrllm/tools/code_tools/lcb_tool.py110 except Exception:CODE
LOWrllm/tools/code_tools/lcb_tool.py203 except Exception as e:CODE
353 more matches not shown…
Unused Imports375 hits · 356 pts
SeverityFileLineSnippetContext
LOWtests/test_verl_metrics.py3CODE
LOWtests/test_store.py3CODE
LOWtests/test_verl_policy_loss.py3CODE
LOWtests/tasks/test_loader_memory.py1CODE
LOWtests/integration/test_sampling_e2e.py10CODE
LOWtests/harnesses/test_cli_harness.py13CODE
LOWtests/cli/test_sampling_config.py3CODE
LOWtests/cli/test_sampling_resolution.py3CODE
LOWtests/sandbox/test_warm_queue.py9CODE
LOWtests/sandbox/test_sandboxed_flow.py3CODE
LOWtests/sandbox/test_snapshot.py10CODE
LOWtests/sandbox/test_daytona_backend.py8CODE
LOWtests/eval/test_unified_runner.py17CODE
LOWtests/eval/test_pull.py3CODE
LOWtests/eval/test_evaluator_loader.py3CODE
LOWtests/eval/test_agent_loader.py3CODE
LOWtests/eval/test_env_join.py12CODE
LOWtests/eval/test_script_module_evaluators.py11CODE
LOWtests/eval/test_agentic.py9CODE
LOWtests/eval/test_vlm_agents.py3CODE
LOWtests/eval/test_eval_engine.py16CODE
LOWtests/eval/test_ifeval.py9CODE
LOWtests/eval/test_rollout_decorator.py3CODE
LOWtests/eval/test_vlm_transforms.py3CODE
LOWtests/eval/test_transforms.py3CODE
LOWtests/eval/test_materialize.py13CODE
LOWtests/eval/test_resolver.py8CODE
LOWtests/eval/test_eval_types.py11CODE
LOWtests/eval/test_mcq.py9CODE
LOWtests/data/test_swesmith_builder.py3CODE
LOWtests/data/test_skillsbench_builder.py8CODE
LOWrllm/hooks.py24CODE
LOWrllm/env.py19CODE
LOWrllm/paths.py18CODE
LOWrllm/types.py18CODE
LOW…erimental/rllm_telemetry/src/rllm_telemetry/scorers.py3CODE
LOW…rllm_telemetry/src/rllm_telemetry/trajectory_export.py19CODE
LOW…perimental/rllm_telemetry/src/rllm_telemetry/config.py3CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/exporter.py13CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py3CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py3CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py5CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py6CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py6CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py6CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py7CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py7CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py7CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py7CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py7CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py7CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py7CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py7CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py7CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py9CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py9CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py9CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py9CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py9CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py9CODE
315 more matches not shown…
Cross-Language Confusion60 hits · 349 pts
SeverityFileLineSnippetContext
HIGHtests/harnesses/test_cli_harness.py98 invocations like ``cd /work && claude ...`` are compound commands andSTRING
HIGHrllm/harnesses/qwen_code.py31 -o Acquire::Check-Valid-Until=false 2>/dev/null || trueCODE
HIGHrllm/harnesses/qwen_code.py92 f". $HOME/.nvm/nvm.sh 2>/dev/null; "CODE
HIGHrllm/harnesses/zeroclaw.py46 apt-get update -qq 2>/dev/null || trueCODE
HIGHrllm/harnesses/zeroclaw.py47 apt-get install -y -qq --no-install-recommends curl ca-certificates tar coreutils 2>/dev/null || trueCODE
HIGHrllm/harnesses/zeroclaw.py146 f"mkdir -p $HOME/.zeroclaw && cat > {_CONFIG_PATH} << 'ZC_CONFIG_EOF'\n{content}\nZC_CONFIG_EOF",CODE
HIGHrllm/harnesses/codex.py44 -o Acquire::Check-Valid-Until=false 2>/dev/null || trueCODE
HIGHrllm/harnesses/codex.py155 f". $HOME/.nvm/nvm.sh 2>/dev/null; "CODE
HIGHrllm/harnesses/kimi_cli.py42 -o Acquire::Check-Valid-Until=false 2>/dev/null || trueCODE
HIGHrllm/harnesses/kimi_cli.py158 f"done; kill 0 2>/dev/null)"CODE
HIGHrllm/harnesses/mini_swe_agent.py55 apt-get update -qq && apt-get install -y -qq curl ca-certificates gitCODE
HIGHrllm/harnesses/mini_swe_agent.py147 f"mkdir -p $HOME/.config/mini-swe-agent && cat > {path} << 'MSWEA_DOTENV_EOF'\n{content}\nMSWEA_DOTENV_EOF",CODE
HIGHrllm/harnesses/aider.py44 -o Acquire::Check-Valid-Until=false 2>/dev/null || trueCODE
HIGHrllm/harnesses/opencode.py47 apt-get update -qq && apt-get install -y -qq curl ca-certificatesCODE
HIGHrllm/harnesses/opencode.py148 cmd = f"mkdir -p $HOME/.config/opencode && cat > $HOME/.config/opencode/opencode.json << '{marker}'\n{content}\nCODE
HIGHrllm/harnesses/opencode.py168 f". $HOME/.nvm/nvm.sh 2>/dev/null; "CODE
HIGHrllm/harnesses/cli_harness.py83 invocations like ``cd /workspace && claude ...`` are compoundSTRING
HIGHrllm/harnesses/terminus2.py63 apt-get update -qq && apt-get install -y -qq tmux util-linux procps curl ca-certificates gitCODE
HIGHrllm/harnesses/claude_code.py65 apt-get "${_APT_OPTS[@]}" update -qq 2>/dev/null || trueCODE
HIGHrllm/harnesses/tools/submit_tool.py31 diff = _safe_exec(sandbox, "git diff 2>/dev/null || echo 'No git repo'")CODE
HIGHrllm/eval/script_evaluator.py119 check = sandbox.exec(f"test -f {path} && echo yes || echo no", timeout=10, user=user).strip()CODE
HIGHrllm/eval/visualizer.py370 if (v == null) return "";CODE
HIGHrllm/eval/visualizer.py372 try { return JSON.stringify(v, null, 2); } catch (e) { return String(v); }CODE
HIGHrllm/eval/visualizer.py375 if (s == null) return "text-gray-400";CODE
HIGHrllm/eval/visualizer.py472 if (av == null && bv == null) return 0;CODE
HIGHrllm/eval/visualizer.py472 if (av == null && bv == null) return 0;CODE
HIGHrllm/eval/visualizer.py473 if (av == null) return 1;CODE
HIGHrllm/eval/visualizer.py474 if (bv == null) return -1;CODE
HIGHrllm/eval/visualizer.py722 if (!task || typeof task !== "object") return fieldBox("task", task);CODE
HIGHrllm/eval/visualizer.py724 const metadata = task.metadata && Object.keys(task.metadata).length ? task.metadata : null;CODE
HIGHrllm/eval/visualizer.py724 const metadata = task.metadata && Object.keys(task.metadata).length ? task.metadata : null;CODE
HIGHrllm/eval/visualizer.py737 for (const [ti, traj] of (ep.trajectories || []).entries()) parts.push(renderTrajectory(traj, ti));CODE
HIGHrllm/eval/visualizer.py738 if (ep.artifacts && Object.keys(ep.artifacts).length) parts.push(fieldBox("artifacts", ep.artifacts));CODE
HIGHrllm/eval/visualizer.py738 if (ep.artifacts && Object.keys(ep.artifacts).length) parts.push(fieldBox("artifacts", ep.artifacts));CODE
HIGHrllm/eval/visualizer.py739 if (ep.metrics && Object.keys(ep.metrics).length) parts.push(fieldBox("metrics", ep.metrics));CODE
HIGHrllm/eval/visualizer.py739 if (ep.metrics && Object.keys(ep.metrics).length) parts.push(fieldBox("metrics", ep.metrics));CODE
HIGHrllm/eval/visualizer.py740 if (ep.metadata && Object.keys(ep.metadata).length) parts.push(fieldBox("metadata", ep.metadata));CODE
HIGHrllm/eval/visualizer.py740 if (ep.metadata && Object.keys(ep.metadata).length) parts.push(fieldBox("metadata", ep.metadata));CODE
HIGHrllm/eval/visualizer.py524let runState = { id:null, meta:null, index:[], filter:"all", search:"" };CODE
HIGHrllm/eval/visualizer.py556 if (summary.total != null) {CODE
HIGHrllm/eval/visualizer.py620 if (r == null) return "";CODE
HIGHrllm/eval/visualizer.py645 if (v == null) return "";CODE
HIGHrllm/eval/visualizer.py677 const meta = step.metadata && Object.keys(step.metadata).length ? step.metadata : null;CODE
HIGHrllm/eval/visualizer.py677 const meta = step.metadata && Object.keys(step.metadata).length ? step.metadata : null;CODE
HIGHrllm/data/r2egym_builder.py291 bash "$RUN_TESTS" > /tmp/test_output.txt 2>&1 || log "run_tests.sh exited non-zero (parser inspects log)"CODE
HIGHrllm/data/r2egym_builder.py296 PYTHONWARNINGS='ignore::UserWarning,ignore::SyntaxWarning' "$PY" -W ignore -m pytest -rA r2e_tests > /tmp/test_outpuCODE
HIGHrllm/data/swebench_pro_builder.py313git add -A . >/dev/null 2>&1 || trueCODE
HIGHrllm/data/swebench_pro_builder.py314git diff --cached --binary "$BASE_COMMIT" > "$MODEL_PATCH" 2>/dev/null || trueCODE
HIGHrllm/data/swebench_pro_builder.py315git reset >/dev/null 2>&1 || trueCODE
HIGHrllm/data/swebench_pro_builder.py317PATCH_BYTES=$(wc -c < "$MODEL_PATCH" 2>/dev/null || echo 0)CODE
HIGHrllm/data/swebench_pro_builder.py334git reset --hard "$BASE_COMMIT" >/dev/null 2>&1 || log "git reset --hard failed (continuing)"CODE
HIGHrllm/data/swebench_pro_builder.py335git checkout "$BASE_COMMIT" >/dev/null 2>&1 || log "git checkout failed (continuing)"CODE
HIGHrllm/data/swebench_pro_builder.py336git clean -fd >/dev/null 2>&1 || log "git clean -fd failed (continuing)"CODE
HIGHrllm/data/swebench_pro_builder.py370bash /tests/run_script.sh "$SELECTED" > /tmp/stdout.log 2> /tmp/stderr.log || log "run_script.sh exited non-zero (parserCODE
HIGHrllm/data/swebench_pro_builder.py373python3 /tests/parser.py /tmp/stdout.log /tmp/stderr.log /tmp/output.json 2>>/tmp/stderr.log || log "parser.py failed"CODE
HIGHrllm/data/swebench_pro_builder.py289git config --global --add safe.directory "$(pwd)" 2>/dev/null || trueCODE
HIGHrllm/data/swebench_pro_builder.py169 - ``requirements``: explicit deliverables (sometimes null)STRING
HIGHrllm/data/swebench_pro_builder.py456 "git config --global --add safe.directory /app 2>/dev/null || true\n"STRING
HIGHrllm/trainer/fireworks/fireworks_backend.py270 raise ValueError(f"rllm.algorithm.loss_agg_mode must be null, 'token-mean', 'seq-mean-token-sum', or 'seq-meCODE
HIGHrllm/trainer/fireworks/fireworks_backend.py278 raise ValueError(f"rollout_correction.tis_mode must be null, 'token', or 'sequence', got '{tis_mode}'")CODE
Deep Nesting169 hits · 157 pts
SeverityFileLineSnippetContext
LOWrllm/hooks.py168CODE
LOWrllm/hooks.py245CODE
LOW…lm/experimental/rllm_telemetry/examples/hello_world.py69CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/exporter.py440CODE
LOW…erimental/rllm_telemetry/src/rllm_telemetry/dataset.py48CODE
LOW…experimental/rllm_telemetry/src/rllm_telemetry/eval.py159CODE
LOW…erimental/rllm_telemetry/src/rllm_telemetry/compare.py88CODE
LOW…erimental/rllm_telemetry/src/rllm_telemetry/compare.py165CODE
LOWrllm/experimental/fully_async/runner.py228CODE
LOWrllm/experimental/fully_async/metric_utils.py122CODE
LOWrllm/experimental/fully_async/rollout_executor.py326CODE
LOWrllm/experimental/fully_async/utils.py251CODE
LOWrllm/experimental/fully_async/utils.py350CODE
LOWrllm/experimental/fully_async/utils.py480CODE
LOWrllm/experimental/fully_async/fully_async_trainer.py166CODE
LOWrllm/experimental/fully_async/fully_async_trainer.py524CODE
LOWrllm/tasks/loader.py169CODE
LOWrllm/tools/utils.py43CODE
LOWrllm/tools/mcp_tool.py20CODE
LOWrllm/tools/code_tools/together_tool.py40CODE
LOWrllm/tools/code_tools/lcb_tool.py58CODE
LOWrllm/tools/code_tools/lcb_tool.py122CODE
LOWrllm/harnesses/tool_calling.py33CODE
LOWrllm/harnesses/tools/file_editor_tool.py61CODE
LOWrllm/rewards/code_reward.py425CODE
LOWrllm/rewards/countdown_reward.py156CODE
LOWrllm/rewards/math_utils/utils.py42CODE
LOWrllm/rewards/math_utils/utils.py496CODE
LOWrllm/rewards/math_utils/utils.py43CODE
LOWrllm/rewards/code_utils/taco.py79CODE
LOWrllm/rewards/code_utils/taco.py214CODE
LOWrllm/rewards/code_utils/taco.py292CODE
LOWrllm/rewards/code_utils/taco.py598CODE
LOWrllm/rewards/code_utils/codeforces.py61CODE
LOWrllm/rewards/code_utils/codeforces.py193CODE
LOWrllm/rewards/code_utils/codeforces.py271CODE
LOWrllm/rewards/code_utils/livecodebench.py268CODE
LOWrllm/rewards/code_utils/livecodebench.py382CODE
LOWrllm/rewards/code_utils/humanevalplus.py16CODE
LOWrllm/workflows/distillation_workflow.py43CODE
LOWrllm/workflows/simple_workflow.py32CODE
LOWrllm/workflows/timing_mixin.py134CODE
LOWrllm/utils/visualization.py69CODE
LOWrllm/utils/tracking.py442CODE
LOWrllm/utils/tracking.py487CODE
LOWrllm/utils/tracking.py608CODE
LOWrllm/cli/_ui.py74CODE
LOWrllm/cli/model_cmd.py35CODE
LOWrllm/cli/dataset.py47CODE
LOWrllm/cli/dataset.py224CODE
LOWrllm/cli/train.py146CODE
LOWrllm/cli/_pull.py153CODE
LOWrllm/cli/_pull.py316CODE
LOWrllm/cli/eval.py80CODE
LOWrllm/cli/eval.py578CODE
LOWrllm/parser/chat_template_parser.py87CODE
LOWrllm/parser/chat_template_parser.py206CODE
LOWrllm/parser/chat_template_parser.py263CODE
LOWrllm/parser/chat_template_parser.py319CODE
LOWrllm/parser/chat_template_parser.py396CODE
109 more matches not shown…
Modern Structural Boilerplate146 hits · 150 pts
SeverityFileLineSnippetContext
LOWrllm/hooks.py42logger = logging.getLogger(__name__)CODE
LOWrllm/hooks.py345__all__ = [CODE
LOWrllm/__init__.py10__all__ = ["BaseAgent", "Action", "Step", "Trajectory", "Episode", "rollout", "evaluator", "Task"]CODE
LOW…rimental/rllm_telemetry/src/rllm_telemetry/__init__.py34__all__ = [CODE
LOWrllm/experimental/fully_async/param_sync.py22logger = logging.getLogger(__name__)CODE
LOWrllm/experimental/fully_async/message_queue.py23logger = logging.getLogger(__name__)CODE
LOWrllm/tasks/__init__.py16__all__ = [CODE
LOWrllm/tasks/loader.py38logger = logging.getLogger(__name__)CODE
LOWrllm/tools/__init__.py25__all__ = ["PythonInterpreter", "GoogleSearchTool", "FirecrawlTool", "TavilyExtractTool", "TavilySearchTool", "ToolRegisCODE
LOWrllm/tools/mcp_tool.py6logger = logging.getLogger(__name__)CODE
LOWrllm/tools/code_tools/__init__.py6__all__ = [CODE
LOWrllm/tools/web_tools/__init__.py5__all__ = ["GoogleSearchTool", "FirecrawlTool", "TavilyExtractTool", "TavilySearchTool"]CODE
LOWrllm/environments/__init__.py3__all__ = ["BaseEnv"]CODE
LOWrllm/environments/base/__init__.py3__all__ = ["BaseEnv"]CODE
LOWrllm/harnesses/react.py19logger = logging.getLogger(__name__)CODE
LOWrllm/harnesses/bash.py23logger = logging.getLogger(__name__)CODE
LOWrllm/harnesses/cli_harness.py41logger = logging.getLogger(__name__)CODE
LOWrllm/harnesses/terminus2.py33logger = logging.getLogger(__name__)CODE
LOWrllm/harnesses/oracle.py34logger = logging.getLogger(__name__)CODE
LOWrllm/harnesses/tool_calling.py17logger = logging.getLogger(__name__)CODE
LOWrllm/harnesses/tools/__init__.py7__all__ = ["BashTool", "FileEditorTool", "SubmitTool"]CODE
LOWrllm/rewards/__init__.py6__all__ = ["RewardInput", "RewardOutput", "RewardType", "RewardConfig", "RewardFunction", "zero_reward"]CODE
LOWrllm/rewards/countdown_reward.py8logger = logging.getLogger(__name__)CODE
LOWrllm/rewards/math_utils/__init__.py11__all__ = ["extract_answer", "grade_answer_sympy", "grade_answer_mathd"]CODE
LOWrllm/rewards/code_utils/pyext2.py5__all__ = [CODE
LOWrllm/workflows/__init__.py8__all__ = [CODE
LOWrllm/agents/__init__.py13__all__ = [CODE
LOWrllm/agents/agent.py29__all__ = [CODE
LOWrllm/utils/__init__.py8__all__ = ["EpisodeLogger", "compute_pass_at_k", "extract_source_metadata", "visualize_trajectories", "VisualizationConfCODE
LOWrllm/cli/_pull.py12logger = logging.getLogger(__name__)CODE
LOWrllm/cli/eval.py24logger = logging.getLogger(__name__)CODE
LOWrllm/parser/chat_template_parser.py10logger = logging.getLogger(__name__)CODE
LOWrllm/parser/__init__.py3__all__ = [CODE
LOWrllm/integrations/harbor/atif_trajectory_bridge.py22logger = logging.getLogger(__name__)CODE
LOWrllm/integrations/harbor/runtime.py24logger = logging.getLogger(__name__)CODE
LOWrllm/integrations/harbor/dataset_loader.py15logger = logging.getLogger(__name__)CODE
LOWrllm/integrations/harbor/evaluator.py18logger = logging.getLogger(__name__)CODE
LOWrllm/integrations/harbor/trial_helper.py21logger = logging.getLogger(__name__)CODE
LOWrllm/sandbox/train_schedule.py55__all__ = ["build_train_schedule"]CODE
LOWrllm/sandbox/sandboxed_flow.py18logger = logging.getLogger(__name__)CODE
LOWrllm/sandbox/warm_queue.py48logger = logging.getLogger(__name__)CODE
LOWrllm/sandbox/warm_queue.py258__all__ = ["WarmQueue"]CODE
LOWrllm/sandbox/snapshot.py26logger = logging.getLogger(__name__)CODE
LOWrllm/sandbox/snapshot.py412__all__ = ["env_key", "env_key_for", "install_script_for", "keys_for_tasks", "get_sandbox", "SnapshotRegistry"]CODE
LOWrllm/sandbox/backends/daytona.py29logger = logging.getLogger(__name__)CODE
LOWrllm/sandbox/backends/modal_backend.py27logger = logging.getLogger(__name__)CODE
LOWrllm/sandbox/backends/local.py11logger = logging.getLogger(__name__)CODE
LOWrllm/sandbox/backends/docker.py10logger = logging.getLogger(__name__)CODE
LOWrllm/eval/module_evaluator.py28logger = logging.getLogger(__name__)CODE
LOWrllm/eval/vlm_utils.py15logger = logging.getLogger(__name__)CODE
LOWrllm/eval/rollout_decorator.py21logger = logging.getLogger(__name__)CODE
LOWrllm/eval/rollout_decorator.py31__all__ = ["_coerce_to_episode"]CODE
LOWrllm/eval/runner.py26logger = logging.getLogger(__name__)CODE
LOWrllm/eval/proxy.py33logger = logging.getLogger(__name__)CODE
LOWrllm/eval/proxy.py192 def set_limits() -> None:CODE
LOWrllm/eval/script_evaluator.py22logger = logging.getLogger(__name__)CODE
LOWrllm/eval/materialize.py39logger = logging.getLogger(__name__)CODE
LOWrllm/eval/_resolution.py37logger = logging.getLogger(__name__)CODE
LOWrllm/eval/_resolution.py329def _setup_task_environment(task: Task, sandbox: Sandbox) -> None:CODE
LOWrllm/eval/reward_fns/llm_equality.py19logger = logging.getLogger(__name__)CODE
86 more matches not shown…
Self-Referential Comments28 hits · 99 pts
SeverityFileLineSnippetContext
MEDIUMtests/test_store.py184 # Create a minimal concrete subclass.COMMENT
MEDIUMtests/test_verl_policy_loss.py116 # Create a mock actor instance with a configCOMMENT
MEDIUMtests/test_verl_policy_loss.py121 # Create a mock DataProto with overrideCOMMENT
MEDIUMtests/data/test_registry_migration.py44 # Create a fake v1 registry at the new locationCOMMENT
MEDIUMtests/data/test_registry_migration.py48 # Create a fake parquet fileCOMMENT
MEDIUMrllm/experimental/fully_async/runner.py291 # Create a configured TaskRunner class with rollout_fn baked inCOMMENT
MEDIUMrllm/tools/__init__.py21# Create the singleton registry instance and register all default toolsCOMMENT
MEDIUMrllm/tools/utils.py161 # Create an underlying bytes buffer initialized with the encoded input.COMMENT
MEDIUMrllm/tools/code_tools/e2b_tool.py83 # Create a CodeToolOutput object instead of a dictionaryCOMMENT
MEDIUMrllm/tools/code_tools/together_tool.py114# Create a StringIO object with some test dataCOMMENT
MEDIUMrllm/tools/code_tools/lcb_tool.py211 # Create a Python interpreter instanceCOMMENT
MEDIUMrllm/rewards/code_reward.py113 # Create a new dict with only the selected test casesCOMMENT
MEDIUMrllm/rewards/countdown_reward.py49 # Define a regex pattern that only allows numbers, operators, parentheses, and whitespaceCOMMENT
MEDIUMrllm/rewards/code_utils/kodcode.py18 # Create a preexec_fn function to set resource limitsCOMMENT
MEDIUMrllm/rewards/code_utils/humanevalplus.py45 # Create a preexec_fn function to set resource limitsCOMMENT
MEDIUMrllm/utils/compute_pass_at_k.py11 # Create a map to store correct answers per problemCOMMENT
MEDIUMrllm/utils/tracking.py865 # Create a new table with same columns and existing dataCOMMENT
MEDIUMrllm/trainer/verl/train_agent_ppo.py38 # Create a remote instance of the TaskRunner class, andCOMMENT
MEDIUMrllm/trainer/verl/train_agent_ppo.py48 # This file is used for performance analysisCOMMENT
MEDIUMexamples/archive/deepresearch/deepresearch_workflow.py54 # Create the DeepResearch agentCOMMENT
MEDIUMexamples/archive/deepresearch/deepresearch_workflow.py103 # Create a failed episodeCOMMENT
MEDIUMexamples/countdown/run_countdown.py53 # Create a map to store correct answers per problemCOMMENT
MEDIUMscripts/data/download_gaia.py7# Define the Google Drive file IDs for the Gaia filesCOMMENT
MEDIUMscripts/data/download_gaia.py16# Define the destination pathsCOMMENT
MEDIUMscripts/data/download_gaia.py22 # Create the necessary directoriesCOMMENT
MEDIUMscripts/data/download_datasets.py8# Define the Google Drive file IDs for the JSON filesCOMMENT
MEDIUMscripts/data/download_datasets.py21# Define the destination pathsCOMMENT
MEDIUMscripts/data/download_datasets.py31# Create the necessary directoriesCOMMENT
Docstring Block Structure14 hits · 70 pts
SeverityFileLineSnippetContext
HIGHrllm/experimental/fully_async/utils.py481 Assemble gen_batch_output from TrajectoryGroup objects. Optimized version: pre-allocates tensors and fills in sSTRING
HIGHrllm/tools/registry.py123 Get a tool class by name using dictionary-like syntax. Args: name: The name of the tool toSTRING
HIGHrllm/environments/base/base_env.py60Creates an environment instance from a dictionary. This method should be implemented by concrete subclasses to STRING
HIGHrllm/parser/chat_template_parser.py51Verify that parsing messages together is equivalent to parsing them individually. Args: messages (lSTRING
HIGHrllm/parser/chat_template_parser.py88Factory method to get the appropriate parser based on a string identifier. Args: parser_type (str):STRING
HIGHrllm/parser/tool_parser.py21Factory method to get the appropriate tool parser based on a string identifier. Args: tokenizer: ThSTRING
HIGHrllm/data/dataset.py136Load dataset directly from a file path. Args: path: Path to the dataset file Returns: STRING
HIGHrllm/data/utils.py43Load a dataset from a JSON file based on the dataset enum. This function takes a dataset enum value and loads the cSTRING
HIGHrllm/engine/unified_workflow_engine.py121Process a single task rollout with retry logic based on termination reasons. Args: task: Task dictiSTRING
HIGHrllm/engine/agent_workflow_engine.py99Process a single task rollout with retry logic based on termination reasons. Args: task: Task dictiSTRING
HIGHrllm/engine/rollout/completer.py25 Basic completer that takes in messages and returns a single step. Args: rollout_engine: The rollout enSTRING
HIGHrllm/trainer/distill/alignment.py211 Align teacher logprobs to student tokens using byte-level alignment. Maps teacher logprobs to student tokens bSTRING
HIGHexamples/fully_async/deepresearch/refine_agent.py311 Parse out the <think>...</think> tags and return only the result part. Args: content: The raw model reSTRING
HIGHexamples/fully_async/deepresearch/refine_agent.py340 Parse out the content inside <information>...</information> tags. Args: content: The model response thSTRING
Cross-File Repetition11 hits · 55 pts
SeverityFileLineSnippetContext
HIGHrllm/rewards/code_utils/taco.py0if test(generated_code) is not none it'll try to run the code. otherwise it'll just return an input and output pair.STRING
HIGHrllm/rewards/code_utils/codeforces.py0if test(generated_code) is not none it'll try to run the code. otherwise it'll just return an input and output pair.STRING
HIGHrllm/rewards/code_utils/livecodebench.py0if test(generated_code) is not none it'll try to run the code. otherwise it'll just return an input and output pair.STRING
HIGHrllm/rewards/code_utils/taco.py0this disables various destructive functions and prevents the generated code from interfering with the test (e.g. fork boSTRING
HIGHrllm/rewards/code_utils/codeforces.py0this disables various destructive functions and prevents the generated code from interfering with the test (e.g. fork boSTRING
HIGHrllm/rewards/code_utils/livecodebench.py0this disables various destructive functions and prevents the generated code from interfering with the test (e.g. fork boSTRING
HIGHrllm/rewards/code_utils/kodcode.py0this disables various destructive functions and prevents the generated code from interfering with the test (e.g. fork boSTRING
HIGHrllm/rewards/code_utils/humanevalplus.py0this disables various destructive functions and prevents the generated code from interfering with the test (e.g. fork boSTRING
HIGHscripts/data/gaia_dataset.py0create a mapping function to process dataset examples. args: split: dataset split name ('train' or 'test') returns: funcSTRING
HIGHscripts/data/code_dataset.py0create a mapping function to process dataset examples. args: split: dataset split name ('train' or 'test') returns: funcSTRING
HIGHscripts/data/deepscaler_dataset.py0create a mapping function to process dataset examples. args: split: dataset split name ('train' or 'test') returns: funcSTRING
AI Slop Vocabulary18 hits · 53 pts
SeverityFileLineSnippetContext
MEDIUMtests/harnesses/test_cli_harness.py258# Cross-harness: ANTHROPIC_BASE_URL must not end in /v1COMMENT
LOWrllm/tools/web_tools/firecrawl_tool.py130 # For now, just call the synchronous versionCOMMENT
MEDIUMrllm/harnesses/zeroclaw.py158 # startup before any LLM call in testing. The symlink is robust.)COMMENT
MEDIUMrllm/harnesses/claude_code.py127 # harness doesn't pollute $HOME between runs.COMMENT
MEDIUMrllm/rewards/code_utils/swebench.py49# from swebench.harness.reporting import make_run_reportCOMMENT
MEDIUMrllm/cli/train.py312 # Harbor datasets on an rllm-native harness skip ``harbor_reward_fn``COMMENT
MEDIUMrllm/cli/eval.py162 # Resolve agent name (display + harness lookup)COMMENT
MEDIUMrllm/cli/eval.py276 # harbor-sourced dataset through an rllm-native harness (oracle,COMMENT
MEDIUMrllm/data/skillsbench_builder.py240# materialization time; the harness-layer hook is the right place for them.COMMENT
MEDIUMrllm/data/swebench_pro_builder.py310# regardless of what the harness or the image left HEAD pointing at.COMMENT
MEDIUMrllm/data/swebench_pro_builder.py648 # solution/ (gold patch + apply script for the oracle harness)COMMENT
LOWrllm/engine/rollout/tinker_engine.py305 sampling_params.pop("max_tokens", None) # TODO(listar2000): later we should simply use this as the max_tokensCODE
LOWrllm/trainer/algorithms/rejection_sampling.py195 # No rejection, just return filtered groups with metricsCOMMENT
LOWrllm/trainer/verl/transform.py184 # The padding is done after the transform (in `_pad_dataproto_to_world_size`), so we simply set all to False herCOMMENT
MEDIUMrllm/trainer/verl/transform.py600 # Match advantages to batch entries by step_id (robust to batch reordering and padding)COMMENT
MEDIUMrllm/trainer/verl/verl_backend.py841 # Async hook methods - leverage RayPPOTrainer utilities where possibleCOMMENT
MEDIUMexamples/fully_async/deepresearch/train.py91 # Be robust: if structure is unexpected, fall back to zeroCOMMENT
MEDIUMexamples/fully_async/deepresearch/train.py147 # Be robust: if structure is unexpected, fall back to zeroCOMMENT
Redundant / Tautological Comments31 hits · 50 pts
SeverityFileLineSnippetContext
LOWrllm/experimental/fully_async/utils.py418 # Check if all correct or all incorrectCOMMENT
LOWrllm/experimental/fully_async/fully_async_trainer.py235 # Check if the ESI (Elastic Server Instance)/training plan is close to expiration.COMMENT
LOWrllm/experimental/fully_async/fully_async_trainer.py240 # Check if the conditions for saving a checkpoint are met.COMMENT
LOWrllm/rewards/code_reward.py62 # Check if we're out of the block (less indentation)COMMENT
LOWrllm/rewards/countdown_reward.py36 # Check if all numbers in equation are availableCOMMENT
LOWrllm/rewards/countdown_reward.py180 # Check if all numbers used are from the available setCOMMENT
LOWrllm/rewards/code_utils/swebench.py203 # Check if git diff changed after running eval scriptCOMMENT
LOWrllm/cli/agent.py45 # Check if it's a plugin agentCOMMENT
LOWrllm/cli/login.py21 # Check if already logged inCOMMENT
LOWrllm/cli/eval.py491 # Print resultsCOMMENT
LOWrllm/parser/chat_template_parser.py73 # Check if results are equivalentCOMMENT
LOWrllm/integrations/harbor/dataset_loader.py176 # Check if it's a local dataset directoryCOMMENT
LOWrllm/trainer/algorithms/rejection_sampling.py201 # Check if we have enough partial solvesCOMMENT
LOWrllm/trainer/verl/agent_workflow_trainer.py228 # Check if all episodes are correct or incorrectCOMMENT
LOWrllm/trainer/verl/train_agent_ppo.py29 # Check if Ray is not initializedCOMMENT
LOWrllm/trainer/deprecated/tinker_data_processor.py466 # Check if current step extends previous sequenceCOMMENT
LOWrllm/trainer/deprecated/tinker_policy_trainer.py74 # Check if a Tinker model ID is provided in configCOMMENT
LOWexamples/archive/sft_tinker/train_norobots_tinker.sh17# Check if TINKER_API_KEY is setCOMMENT
LOWexamples/archive/deepresearch/deepresearch_tools.py634 # Check if the module or its parent is allowedCOMMENT
LOWexamples/archive/deepresearch/deepresearch_agent.py568 # Check if we've exceeded call limitCOMMENT
LOWexamples/archive/deepresearch/evaluate_hle.py355 # Check if model is O3/O1 (use model_name which is already determined above)COMMENT
LOWexamples/fully_async/deepresearch/refine_agent.py326 # Check if there's a <think> tagCOMMENT
LOWexamples/fully_async/deepresearch/refine_agent.py330 # Check if there's a </think> tagCOMMENT
LOWexamples/fully_async/deepresearch/refine_agent.py355 # Check if there's an <information> tagCOMMENT
LOWexamples/fully_async/deepresearch/refine_agent.py359 # Check if there's a </information> tagCOMMENT
LOWexamples/fully_async/deepresearch/search_agent.py338 # Check if duplicate search was detected - if so, set reward to 0COMMENT
LOWexamples/fully_async/deepresearch/search_agent.py347 # Check if excessive parallel tool calls detected (>= 5) - if so, set reward to 0COMMENT
LOWexamples/fully_async/deepresearch/tool.py116 self._maybe_refresh_urls() # Check if URLs need refreshingCODE
LOWexamples/fully_async/deepresearch/rag/launch_rag.sh61# Check if data directory existsCOMMENT
LOWscripts/benchmark/cf_elo_calc.py212 # Display resultsCOMMENT
LOWcookbooks/workflow/solver_judge/solver_judge_flow.py118 # Assign rewards to solver trajectoriesCOMMENT
Verbosity Indicators25 hits · 49 pts
SeverityFileLineSnippetContext
LOWtests/unified_trainer/test_tinker_transform.py280 # Step 1: prompt=[1,2], response=[3,4]COMMENT
LOWtests/unified_trainer/test_tinker_transform.py287 # Step 2: prompt=[1,2,3,4,5] (extends [1,2,3,4]), response=[6,7]COMMENT
LOWtests/unified_trainer/test_tinker_transform.py418 # Step 1: [img_chunk, 1, 2] -> response [3, 4]COMMENT
LOWtests/unified_trainer/test_tinker_transform.py424 # Step 2: extends with [img_chunk, 1, 2, 3, 4, 5] -> response [6, 7]COMMENT
LOWrllm/data/swebench_pro_builder.py303# Step 1: capture the agent's diff vs base_commit.COMMENT
LOWrllm/data/swebench_pro_builder.py320# Step 2: reset to base_commit so the patch applies on a clean tree.COMMENT
LOWrllm/data/swebench_pro_builder.py338# Step 3: re-apply the agent's patch.COMMENT
LOWrllm/data/swebench_pro_builder.py347# Step 4: run the last non-blank line of before_repo_set_cmd. UpstreamCOMMENT
LOWrllm/data/swebench_pro_builder.py362# Step 5: run the per-instance run_script.sh with the selected test files.COMMENT
LOWrllm/data/swebench_pro_builder.py372# Step 6: parse the run_script.sh output into the standard JSON shape.COMMENT
LOWrllm/data/swebench_pro_builder.py375# Step 7: compare passed set against fail_to_pass ∪ pass_to_pass.COMMENT
LOWrllm/trainer/algorithms/rejection_sampling.py186 # Step 1: Filter groups and episodes based on configCOMMENT
LOWrllm/trainer/algorithms/rejection_sampling.py190 # Step 2: Compute episode-level correctness metrics (always, for logging)COMMENT
LOWrllm/trainer/algorithms/rejection_sampling.py193 # Step 3: Apply mode-specific logic (TODO(listar2000): implement a group-level rejection sampling)COMMENT
LOWrllm/trainer/algorithms/transform.py239 # Step 1: Name imputationCOMMENT
LOWrllm/trainer/algorithms/transform.py247 # Step 2: Invoke the trajectory grouping hookCOMMENT
LOWrllm/trainer/algorithms/transform.py250 # Step 3: Get metricsCOMMENT
LOWrllm/trainer/deprecated/tinker_policy_trainer.py228 # Step 1: Process to datums (includes filtering and advantage computation)COMMENT
LOWrllm/trainer/deprecated/tinker_policy_trainer.py236 # Step 3: Remove mask from datums (not needed by forward_backward)COMMENT
LOWrllm/trainer/deprecated/tinker_policy_trainer.py239 # Step 4: Forward-backward passCOMMENT
LOWrllm/trainer/deprecated/tinker_policy_trainer.py245 # Step 5: Optimizer stepCOMMENT
LOWcookbooks/workflow/solver_judge/solver_judge_flow.py115 # Step 1: Solver generates multiple solutions in parallelCOMMENT
LOWcookbooks/workflow/solver_judge/solver_judge_flow.py127 # Step 2: Judge selects the best solutionCOMMENT
LOWcookbooks/solver_judge_flow/solver_judge_flow.py27 # Step 1: Solver generates N solutions in parallelCOMMENT
LOWcookbooks/solver_judge_flow/solver_judge_flow.py30 # Step 2: Judge selects the best solutionCOMMENT
Structural Annotation Overuse24 hits · 48 pts
SeverityFileLineSnippetContext
LOWtests/unified_trainer/test_tinker_transform.py280 # Step 1: prompt=[1,2], response=[3,4]COMMENT
LOWtests/unified_trainer/test_tinker_transform.py287 # Step 2: prompt=[1,2,3,4,5] (extends [1,2,3,4]), response=[6,7]COMMENT
LOWtests/unified_trainer/test_tinker_transform.py424 # Step 2: extends with [img_chunk, 1, 2, 3, 4, 5] -> response [6, 7]COMMENT
LOWrllm/data/swebench_pro_builder.py303# Step 1: capture the agent's diff vs base_commit.COMMENT
LOWrllm/data/swebench_pro_builder.py320# Step 2: reset to base_commit so the patch applies on a clean tree.COMMENT
LOWrllm/data/swebench_pro_builder.py338# Step 3: re-apply the agent's patch.COMMENT
LOWrllm/data/swebench_pro_builder.py347# Step 4: run the last non-blank line of before_repo_set_cmd. UpstreamCOMMENT
LOWrllm/data/swebench_pro_builder.py362# Step 5: run the per-instance run_script.sh with the selected test files.COMMENT
LOWrllm/data/swebench_pro_builder.py372# Step 6: parse the run_script.sh output into the standard JSON shape.COMMENT
LOWrllm/data/swebench_pro_builder.py375# Step 7: compare passed set against fail_to_pass ∪ pass_to_pass.COMMENT
LOWrllm/trainer/algorithms/rejection_sampling.py186 # Step 1: Filter groups and episodes based on configCOMMENT
LOWrllm/trainer/algorithms/rejection_sampling.py190 # Step 2: Compute episode-level correctness metrics (always, for logging)COMMENT
LOWrllm/trainer/algorithms/rejection_sampling.py193 # Step 3: Apply mode-specific logic (TODO(listar2000): implement a group-level rejection sampling)COMMENT
LOWrllm/trainer/algorithms/transform.py239 # Step 1: Name imputationCOMMENT
LOWrllm/trainer/algorithms/transform.py247 # Step 2: Invoke the trajectory grouping hookCOMMENT
LOWrllm/trainer/algorithms/transform.py250 # Step 3: Get metricsCOMMENT
LOWrllm/trainer/deprecated/tinker_policy_trainer.py228 # Step 1: Process to datums (includes filtering and advantage computation)COMMENT
LOWrllm/trainer/deprecated/tinker_policy_trainer.py236 # Step 3: Remove mask from datums (not needed by forward_backward)COMMENT
LOWrllm/trainer/deprecated/tinker_policy_trainer.py239 # Step 4: Forward-backward passCOMMENT
LOWrllm/trainer/deprecated/tinker_policy_trainer.py245 # Step 5: Optimizer stepCOMMENT
LOWcookbooks/workflow/solver_judge/solver_judge_flow.py115 # Step 1: Solver generates multiple solutions in parallelCOMMENT
LOWcookbooks/workflow/solver_judge/solver_judge_flow.py127 # Step 2: Judge selects the best solutionCOMMENT
LOWcookbooks/solver_judge_flow/solver_judge_flow.py27 # Step 1: Solver generates N solutions in parallelCOMMENT
LOWcookbooks/solver_judge_flow/solver_judge_flow.py30 # Step 2: Judge selects the best solutionCOMMENT
Over-Commented Block42 hits · 40 pts
SeverityFileLineSnippetContext
LOWtests/tools/test_tools.py121 "google_search": [COMMENT
LOWrllm/experimental/fully_async/runner.py1# Copyright 2025 Meituan Ltd. and/or its affiliatesCOMMENT
LOWrllm/experimental/fully_async/metric_utils.py1# Copyright 2025 Meituan Ltd. and/or its affiliatesCOMMENT
LOWrllm/experimental/fully_async/param_sync.py1# Copyright 2025 Meituan Ltd. and/or its affiliatesCOMMENT
LOWrllm/experimental/fully_async/message_queue.py1# Copyright 2025 Meituan Ltd. and/or its affiliatesCOMMENT
LOWrllm/experimental/fully_async/inference_manager.py1# Copyright 2025 Meituan Ltd. and/or its affiliatesCOMMENT
LOWrllm/experimental/fully_async/fully_async_trainer.py1# Copyright 2025 Meituan Ltd. and/or its affiliatesCOMMENT
LOWrllm/experimental/fully_async/fully_async_trainer.py241 # The conditions include a mandatory condition (1) andCOMMENT
LOWrllm/experimental/fully_async/fully_async_trainer.py601 if rollout_corr_config is not None and "rollout_log_probs" in batch.batch:COMMENT
LOWrllm/harnesses/zeroclaw.py141 content = self._config_toml(config)COMMENT
LOWrllm/harnesses/kimi_cli.py121 self._exec_agent(sandbox, self._heredoc_write(_KIMI_CONFIG_PATH, content), env=env)COMMENT
LOWrllm/harnesses/claude_code.py21"""COMMENT
LOWrllm/harnesses/claude_code.py121 "ANTHROPIC_MODEL": model,COMMENT
LOWrllm/rewards/code_utils/taco.py421 # result = subprocess.run(['python', temp_program_path], stdin=inputs_tmp_file, text=True, captuCOMMENT
LOWrllm/rewards/code_utils/codeforces.py361 exec_code = -2COMMENT
LOWrllm/utils/tracking.py1# Copyright 2024 Bytedance Ltd. and/or its affiliatesCOMMENT
LOWrllm/cli/eval.py261 # catalog reward_fn > None. When ``evaluator`` is None ``SandboxTaskHooks``COMMENT
LOWrllm/data/r2egym_builder.py261# R2E-Gym's grader runs the repo's own ``run_tests.sh`` from the repo root.COMMENT
LOWrllm/data/skillsbench_builder.py221_SKILLS_COPY_MARKER = "# rllm-skillsbench: skills injection"COMMENT
LOWrllm/data/swebench_pro_builder.py321#COMMENT
LOWrllm/trainer/tinker/transform.py161 datums = []COMMENT
LOWrllm/trainer/fireworks/fireworks_backend.py21from training.provision import FireworksProvisionInfra, init_fireworks_infraCOMMENT
LOWrllm/trainer/algorithms/advantage.py241COMMENT
LOWrllm/trainer/algorithms/advantage.py261 # rewards bucketed sensibly rather than silently dropped.COMMENT
LOWrllm/trainer/verl/transform.py301 # ------------------------------------------------------------------COMMENT
LOWrllm/trainer/sft/config/fireworks.yaml1# Native config for the Fireworks SFT backend.COMMENT
LOWexamples/archive/sft_tinker/train_norobots_tinker.sh1#!/bin/bashCOMMENT
LOWexamples/fully_async/deepresearch/rag/launch_rag.sh1#!/bin/bashCOMMENT
LOWexamples/verifiers_env/train_verifiers_tinker.sh1#!/bin/bashCOMMENT
LOW.github/workflows/claude.yml41COMMENT
LOWcookbooks/math/train_fireworks.sh1#!/usr/bin/env bashCOMMENT
LOWcookbooks/frozenlake/train_fireworks.sh1#!/usr/bin/env bashCOMMENT
LOWcookbooks/deepcoder/train_fireworks.sh1#!/usr/bin/env bashCOMMENT
LOWcookbooks/agent_frameworks/train_fireworks.sh1#!/usr/bin/env bashCOMMENT
LOWcookbooks/agent_frameworks/train_tinker.sh1#!/usr/bin/env bashCOMMENT
LOWcookbooks/agent_frameworks/train_verl.sh1#!/usr/bin/env bashCOMMENT
LOWcookbooks/finqa/train_fireworks.sh1#!/usr/bin/env bashCOMMENT
LOWcookbooks/workflow/solver_judge/train_fireworks.sh1#!/usr/bin/env bashCOMMENT
LOWcookbooks/workflow/geo3k/train_fireworks.sh1#!/usr/bin/env bashCOMMENT
LOWcookbooks/geo3k/train_fireworks.sh1#!/usr/bin/env bashCOMMENT
LOWcookbooks/math_tool_agent/train_fireworks.sh1#!/usr/bin/env bashCOMMENT
LOWcookbooks/solver_judge_flow/train_fireworks.sh1#!/usr/bin/env bashCOMMENT
AI Structural Patterns29 hits · 28 pts
SeverityFileLineSnippetContext
LOWtests/test_store.py134CODE
LOWrllm/cli/eval.py80CODE
LOWrllm/integrations/harbor/runtime.py44CODE
LOWrllm/integrations/harbor/trial_helper.py107CODE
LOWrllm/integrations/harbor/trial_helper.py303CODE
LOWrllm/eval/runner.py29CODE
LOWrllm/data/claw_eval_builder.py92CODE
LOWrllm/data/r2egym_builder.py532CODE
LOWrllm/data/skillsbench_builder.py372CODE
LOWrllm/data/swebench_pro_builder.py667CODE
LOWrllm/data/deepswe_builder.py99CODE
LOWrllm/data/swesmith_builder.py160CODE
LOWrllm/engine/rollout/openai_engine.py21CODE
LOWrllm/engine/rollout/fireworks_engine.py89CODE
LOWrllm/engine/rollout/tinker_engine.py150CODE
LOWrllm/trainer/unified_trainer.py123CODE
LOWrllm/trainer/unified_trainer.py969CODE
LOWrllm/trainer/distill/advantage.py77CODE
LOWexamples/archive/ocr/process_ocr_data.py24CODE
LOWexamples/fully_async/deepresearch/refine_agent.py336CODE
LOWexamples/fully_async/deepresearch/refine_agent.py371CODE
LOWexamples/fully_async/deepresearch/search_agent.py68CODE
LOWexamples/fully_async/deepresearch/tool.py15CODE
LOWexamples/fully_async/deepresearch/rag/rag_server.py59CODE
LOWscripts/data/webarena_dataset.py56CODE
LOWscripts/data/miniwob_dataset.py57CODE
LOWscripts/data/gaia_dataset.py59CODE
LOWscripts/data/code_dataset.py67CODE
LOWscripts/data/deepscaler_dataset.py66CODE
Modern AI Meta-Vocabulary10 hits · 28 pts
SeverityFileLineSnippetContext
MEDIUM…rllm_telemetry/src/rllm_telemetry/trajectory_export.py11 # After the agentic loop finishes:STRING
MEDIUMrllm/cli/eval.py114 # whenever --agent is set to anything other than a harbor scaffoldCOMMENT
MEDIUMrllm/cli/eval.py251 # plus user-registered + plugin agents; ``harbor:<scaffold>`` resolvesCOMMENT
MEDIUMrllm/integrations/harbor/trial_helper.py164 # agent scaffolds read. Each scaffold checks a different one:COMMENT
MEDIUMrllm/engine/remote_runtime/protocol.py92 # Pass-through to harbor AgentConfig.kwargs (scaffold-specific flags).COMMENT
MEDIUMrllm/trainer/config/rllm/base.yaml140 agent: mini-swe-agent # scaffold name (mini-swe-agent | aider | openhands | swe-agent | qwen-coder | ...)CODE
MEDIUMrllm/trainer/config/rllm/base.yaml142 agent_kwargs: {} # pass-through to harbor's AgentConfig.kwargs (scaffold-specific flags)CODE
MEDIUMrllm/trainer/algorithms/advantage.py300 # but role-keyed for multi-agent correctness (single-agent -> "default").COMMENT
MEDIUMrllm/trainer/verl/verl_backend.py446 # and can grow up to the full context window - so use max_total_length toCOMMENT
MEDIUMrllm/trainer/deprecated/tinker_data_processor.py705 # Group by (task_id, trajectory_name) - for multi-agent workflows like solver-judgeCOMMENT
Magic Placeholder Names4 hits · 18 pts
SeverityFileLineSnippetContext
HIGHexamples/archive/sft_tinker/README_NOROBOTS.md162export TINKER_API_KEY=your_api_keyCODE
HIGHexamples/archive/sft_tinker/train_norobots_tinker.sh20 echo "Please set it with: export TINKER_API_KEY=your_api_key"CODE
HIGHexamples/fully_async/deepresearch/search_agent.py8 export OPENAI_API_KEY="your-api-key"STRING
HIGHexamples/verifiers_env/README.md77echo 'TINKER_API_KEY="your-api-key"' > examples/verifiers_env/.envCODE
Example Usage Blocks6 hits · 9 pts
SeverityFileLineSnippetContext
LOWrllm/data/utils.py111 # Example usageCOMMENT
LOWexamples/fully_async/deepresearch/refine_agent.py501 # Example usageCOMMENT
LOWexamples/fully_async/deepresearch/rag/launch_rag.sh8# Usage:COMMENT
LOWcookbooks/agent_frameworks/train_fireworks.sh4# Usage:COMMENT
LOWcookbooks/agent_frameworks/train_tinker.sh4# Usage:COMMENT
LOWcookbooks/agent_frameworks/train_verl.sh4# Usage:COMMENT
Slop Phrases2 hits · 4 pts
SeverityFileLineSnippetContext
LOWrllm/trainer/verl/dataclass.py93 if step_data.advantage is not None: # make sure to not add None to the listCODE
MEDIUMrllm/trainer/verl/train_agent_ppo.py111 # TODO Here you can use the new registration method to support dynamic registration of rolesCOMMENT
TODO Padding2 hits · 3 pts
SeverityFileLineSnippetContext
LOWrllm/engine/rollout/verl_engine.py140 # TODO: implement parse_completion for the standard parserCOMMENT
LOWrllm/trainer/verl/agent_workflow_trainer.py530 # TODO: implement actual tflpo and theoretical tflpoCOMMENT
Overly Generic Function Names1 hit · 2 pts
SeverityFileLineSnippetContext
LOWexamples/fully_async/deepresearch/search_agent.py105 async def helper(tool_call):CODE
Fake / Example Data1 hit · 1 pts
SeverityFileLineSnippetContext
LOWrllm/data/dataset.py730 "prompt": [{"role": "user", "content": "placeholder"}],CODE