# Concepts

Nouns map atomically to file sets (EXTRACTED); verbs aggregate structural edges (INFERRED).

- `topo` | files=27 | mentions=99 | `app.py`, `eval/diag_static.py`, `eval/harness.py`, `eval/hodge_cm_ablation.py`, `eval/noise_sweep.py`, `eval/smoke.py`, `infer_exploitgym.py`, `run_exploitgym.sh`, `synthetic_dataset.py`, `tests/test_jlens.py`
- `model` | files=25 | mentions=186 | `eval/analyze_results.py`, `eval/governor.py`, `eval/governor_smoke.py`, `eval/harness.py`, `eval/noise_sweep.py`, `eval/repair.py`, `eval/sandbox.py`, `infer_exploitgym.py`, `run_exploitgym.sh`, `synthetic_dataset.py`
- `eval` | files=25 | mentions=77 | `eval/analyze.py`, `eval/analyze_results.py`, `eval/diag_static.py`, `eval/governor.py`, `eval/governor_smoke.py`, `eval/harness.py`, `eval/hodge_cm_ablation.py`, `eval/integration_smoke.py`, `eval/noise_analysis.py`, `eval/noise_sweep.py`
- `run` | files=22 | mentions=70 | `app.py`, `eval/analyze_results.py`, `eval/governor.py`, `eval/harness.py`, `eval/integration_smoke.py`, `eval/repair.py`, `eval/report.py`, `eval/samplers.py`, `eval/sandbox.py`, `eval/smoke.py`
- `gpt3` | files=20 | mentions=48 | `app.py`, `eval/diag_static.py`, `eval/harness.py`, `eval/hodge_cm_ablation.py`, `eval/noise_sweep.py`, `eval/smoke.py`, `tests/test_jlens.py`, `tests/test_lens_model.py`, `topogpt3/__init__.py`, `topogpt3/__main__.py`
- `topogpt3` | files=20 | mentions=38 | `app.py`, `eval/governor.py`, `eval/harness.py`, `topogpt3/__init__.py`, `topogpt3/__main__.py`, `topogpt3/api_server.py`, `topogpt3/continuation.py`, `topogpt3/ewc.py`, `topogpt3/exploitgym_config.py`, `topogpt3/exploitgym_loader.py`
- `load` | files=19 | mentions=40 | `eval/analyze.py`, `eval/analyze_results.py`, `eval/governor_smoke.py`, `eval/harness.py`, `eval/noise_analysis.py`, `eval/noise_sweep.py`, `eval/report.py`, `eval/smoke.py`, `infer_exploitgym.py`, `synthetic_dataset.py`
- `build` | files=18 | mentions=44 | `app.py`, `eval/harness.py`, `eval/repair.py`, `eval/samplers.py`, `eval/sandbox.py`, `infer_exploitgym.py`, `run_hodge_cm_ablation.sh`, `synthetic_dataset.py`, `tests/test_lens_model.py`, `topogpt3/api_server.py`
- `prompt` | files=16 | mentions=64 | `eval/analyze_results.py`, `eval/governor.py`, `eval/harness.py`, `eval/integration_smoke.py`, `eval/noise_analysis.py`, `eval/repair.py`, `infer_exploitgym.py`, `synthetic_dataset.py`, `tests/test_jlens.py`, `topogpt3/__init__.py`
- `returns` | files=16 | mentions=53 | `eval/governor.py`, `eval/harness.py`, `eval/samplers.py`, `eval/sandbox.py`, `synthetic_dataset.py`, `tests/test_jlens.py`, `tests/test_lens_model.py`, `topogpt3/continuation.py`, `topogpt3/ewc.py`, `topogpt3/inference.py`
- `when` | files=16 | mentions=29 | `app.py`, `eval/analyze.py`, `eval/governor.py`, `eval/hodge_cm_ablation.py`, `eval/samplers.py`, `eval/sandbox.py`, `infer_exploitgym.py`, `tests/test_lens_model.py`, `topogpt3/api_server.py`, `topogpt3/continuation.py`
- `checkpoint` | files=15 | mentions=67 | `app.py`, `eval/diag_static.py`, `eval/hodge_cm_ablation.py`, `eval/noise_sweep.py`, `eval/smoke.py`, `tests/test_jlens.py`, `tests/test_lens_model.py`, `topogpt3/__init__.py`, `topogpt3/api_server.py`, `topogpt3/inference.py`
- `config` | files=15 | mentions=58 | `eval/sandbox.py`, `eval/temp_sweep.py`, `tests/test_jlens.py`, `tests/test_lens_model.py`, `topogpt3/api_server.py`, `topogpt3/exploitgym_config.py`, `topogpt3/exploitgym_loader.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`, `topogpt3/jlens.py`
- `one` | files=15 | mentions=27 | `eval/analyze.py`, `eval/governor_smoke.py`, `eval/harness.py`, `eval/integration_smoke.py`, `eval/noise_sweep.py`, `eval/samplers.py`, `eval/sandbox.py`, `eval/temp_sweep.py`, `synthetic_dataset.py`, `tests/test_jlens.py`
- `all` | files=14 | mentions=37 | `eval/governor.py`, `eval/hodge_cm_ablation.py`, `eval/integration_smoke.py`, `eval/sandbox_smoke.py`, `tests/test_jlens.py`, `tests/test_lens_model.py`, `topogpt3/ewc.py`, `topogpt3/exploitgym_loader.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`
- `new` | files=14 | mentions=28 | `eval/governor.py`, `eval/harness.py`, `eval/integration_smoke.py`, `eval/repair.py`, `eval/samplers.py`, `topogpt3/__init__.py`, `topogpt3/ewc.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`, `topogpt3/jlens.py`
- `runs` | files=14 | mentions=24 | `eval/analyze_results.py`, `eval/diag_static.py`, `eval/harness.py`, `eval/hodge_cm_ablation.py`, `eval/noise_analysis.py`, `eval/noise_sweep.py`, `eval/repair.py`, `eval/report.py`, `eval/sandbox.py`, `eval/temp_sweep.py`
- `layer` | files=13 | mentions=83 | `eval/hodge_cm_ablation.py`, `eval/sandbox.py`, `tests/test_jlens.py`, `tests/test_lens_model.py`, `topogpt3/api_server.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`, `topogpt3/jlens.py`, `topogpt3/lens_model.py`, `topogpt3/model.py`
- `per` | files=13 | mentions=52 | `eval/analyze.py`, `eval/governor.py`, `eval/governor_smoke.py`, `eval/noise_analysis.py`, `eval/sandbox.py`, `synthetic_dataset.py`, `topogpt3/api_server.py`, `topogpt3/ewc.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`
- `pass` | files=13 | mentions=37 | `eval/analyze.py`, `eval/harness.py`, `eval/integration_smoke.py`, `eval/noise_analysis.py`, `eval/noise_sweep.py`, `eval/report.py`, `eval/temp_sweep.py`, `tests/test_lens_model.py`, `topogpt3/api_server.py`, `topogpt3/ewc.py`
- `code` | files=13 | mentions=30 | `eval/governor.py`, `eval/harness.py`, `eval/samplers.py`, `eval/sandbox.py`, `run_hodge_cm_ablation.sh`, `synthetic_dataset.py`, `topogpt3/__init__.py`, `topogpt3/exploitgym_loader.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`
- `token` | files=12 | mentions=88 | `eval/governor.py`, `eval/governor_smoke.py`, `eval/harness.py`, `eval/noise_analysis.py`, `infer_exploitgym.py`, `tests/test_lens_model.py`, `topogpt3/api_server.py`, `topogpt3/inference_hrm.py`, `topogpt3/jlens.py`, `topogpt3/lens_model.py`
- `file` | files=12 | mentions=57 | `app.py`, `synthetic_dataset.py`, `tests/test_jlens.py`, `tests/test_lens_model.py`, `topogpt3/ewc.py`, `topogpt3/exploitgym_loader.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`, `topogpt3/jlens.py`, `topogpt3/lens_model.py`
- `return` | files=12 | mentions=49 | `app.py`, `eval/governor.py`, `eval/harness.py`, `eval/sandbox.py`, `infer_exploitgym.py`, `tests/test_jlens.py`, `topogpt3/api_server.py`, `topogpt3/continuation.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`
- `error` | files=12 | mentions=34 | `eval/analyze.py`, `eval/harness.py`, `eval/report.py`, `eval/sandbox_smoke.py`, `tests/test_jlens.py`, `tests/test_lens_model.py`, `topogpt3/api_server.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`, `topogpt3/jlens.py`
- `same` | files=12 | mentions=15 | `app.py`, `eval/governor.py`, `eval/harness.py`, `eval/repair.py`, `eval/sandbox.py`, `tests/test_jlens.py`, `tests/test_lens_model.py`, `topogpt3/__init__.py`, `topogpt3/ewc.py`, `topogpt3/jlens.py`
- `max` | files=11 | mentions=37 | `eval/diag_static.py`, `eval/governor.py`, `eval/sandbox.py`, `tests/test_lens_model.py`, `topogpt3/__init__.py`, `topogpt3/api_server.py`, `topogpt3/ewc.py`, `topogpt3/inference_hrm.py`, `topogpt3/jlens.py`, `topogpt3/model.py`
- `top` | files=11 | mentions=36 | `app.py`, `eval/governor.py`, `eval/harness.py`, `eval/repair.py`, `eval/sandbox.py`, `eval/temp_sweep.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`, `topogpt3/jlens.py`, `topogpt3/model.py`
- `each` | files=11 | mentions=30 | `eval/analyze_results.py`, `eval/governor.py`, `eval/harness.py`, `eval/repair.py`, `eval/sandbox.py`, `synthetic_dataset.py`, `topogpt3/ewc.py`, `topogpt3/exploitgym_loader.py`, `topogpt3/inference_hrm.py`, `topogpt3/jlens.py`
- `train` | files=11 | mentions=28 | `run_exploitgym.sh`, `run_exploitgym_v2.sh`, `run_hodge_cm_ablation.sh`, `run_merged.sh`, `topogpt3/__init__.py`, `topogpt3/ewc.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`, `topogpt3/model.py`, `topogpt3/train.py`
- `output` | files=11 | mentions=19 | `eval/harness.py`, `eval/repair.py`, `eval/samplers.py`, `eval/sandbox.py`, `eval/sandbox_smoke.py`, `infer_exploitgym.py`, `tests/test_lens_model.py`, `topogpt3/api_server.py`, `topogpt3/inference_hrm.py`, `topogpt3/jlens.py`
- `usage` | files=11 | mentions=13 | `app.py`, `eval/analyze_results.py`, `eval/governor.py`, `eval/sandbox.py`, `run_exploitgym.sh`, `run_hodge_cm_ablation.sh`, `topogpt3/__init__.py`, `topogpt3/api_server.py`, `topogpt3/ewc.py`, `topogpt3/model.py`
- `generation` | files=10 | mentions=35 | `eval/analyze.py`, `eval/governor.py`, `eval/governor_smoke.py`, `eval/harness.py`, `eval/noise_analysis.py`, `infer_exploitgym.py`, `topogpt3/continuation.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`, `topogpt3/model.py`
- `gpt2` | files=10 | mentions=31 | `eval/noise_sweep.py`, `synthetic_dataset.py`, `tests/test_lens_model.py`, `topogpt3/__init__.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`, `topogpt3/lens_model.py`, `topogpt3/model.py`, `topogpt3/train.py`, `topogpt3/vanilla_control.py`
- `text` | files=10 | mentions=31 | `app.py`, `eval/governor.py`, `eval/harness.py`, `infer_exploitgym.py`, `tests/test_lens_model.py`, `topogpt3/api_server.py`, `topogpt3/continuation.py`, `topogpt3/jlens.py`, `topogpt3/lens_model.py`, `topogpt3/model.py`
- `data` | files=10 | mentions=25 | `run_exploitgym.sh`, `run_merged.sh`, `synthetic_dataset.py`, `topogpt3/ewc.py`, `topogpt3/exploitgym_config.py`, `topogpt3/exploitgym_loader.py`, `topogpt3/jlens.py`, `topogpt3/merged_config.py`, `topogpt3/model.py`, `topogpt3/train.py`
- `training` | files=10 | mentions=25 | `app.py`, `eval/harness.py`, `topogpt3/ewc.py`, `topogpt3/exploitgym_config.py`, `topogpt3/exploitgym_loader.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`, `topogpt3/lens_model.py`, `topogpt3/model.py`, `topogpt3/vanilla_control.py`
- `weights` | files=10 | mentions=25 | `eval/hodge_cm_ablation.py`, `infer_exploitgym.py`, `run_exploitgym.sh`, `tests/test_lens_model.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`, `topogpt3/jlens.py`, `topogpt3/lens_model.py`, `topogpt3/model.py`, `transfer_weights.py`
- `single` | files=10 | mentions=24 | `eval/analyze.py`, `eval/governor.py`, `eval/harness.py`, `infer_exploitgym.py`, `synthetic_dataset.py`, `tests/test_lens_model.py`, `topogpt3/ewc.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`, `topogpt3/model.py`
- `shape` | files=10 | mentions=23 | `eval/repair.py`, `tests/test_jlens.py`, `tests/test_lens_model.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`, `topogpt3/jlens.py`, `topogpt3/lens_model.py`, `topogpt3/model.py`, `topogpt3/train.py`, `transfer_weights.py`
- `full` | files=10 | mentions=21 | `app.py`, `run_exploitgym.sh`, `synthetic_dataset.py`, `tests/test_lens_model.py`, `topogpt3/ewc.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`, `topogpt3/jlens.py`, `topogpt3/model.py`, `topogpt3/vanilla_control.py`
- `human` | files=10 | mentions=20 | `eval/analyze.py`, `eval/analyze_results.py`, `eval/harness.py`, `eval/integration_smoke.py`, `eval/noise_sweep.py`, `eval/repair.py`, `eval/samplers.py`, `eval/sandbox.py`, `eval/smoke.py`, `eval/temp_sweep.py`
- `true` | files=10 | mentions=20 | `eval/governor.py`, `eval/integration_smoke.py`, `infer_exploitgym.py`, `synthetic_dataset.py`, `topogpt3/continuation.py`, `topogpt3/inference_hrm.py`, `topogpt3/jlens.py`, `topogpt3/merged_config.py`, `topogpt3/model.py`, `topogpt3/train.py`
- `loader` | files=10 | mentions=19 | `eval/harness.py`, `eval/noise_sweep.py`, `eval/repair.py`, `topogpt3/exploitgym_config.py`, `topogpt3/exploitgym_loader.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`, `topogpt3/merged_config.py`, `topogpt3/model.py`, `topogpt3/train.py`
- `not` | files=10 | mentions=19 | `eval/governor.py`, `eval/harness.py`, `eval/samplers.py`, `eval/sandbox.py`, `tests/test_jlens.py`, `tests/test_lens_model.py`, `topogpt3/ewc.py`, `topogpt3/inference_hrm.py`, `topogpt3/jlens.py`, `topogpt3/lens_model.py`
- `tensor` | files=10 | mentions=19 | `eval/governor.py`, `tests/test_lens_model.py`, `topogpt3/ewc.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`, `topogpt3/jlens.py`, `topogpt3/lens_model.py`, `topogpt3/model.py`, `topogpt3/train.py`, `transfer_weights.py`
- `real` | files=10 | mentions=15 | `eval/diag_static.py`, `eval/governor.py`, `eval/harness.py`, `eval/hodge_cm_ablation.py`, `topogpt3/api_server.py`, `topogpt3/hodge_cm.py`, `topogpt3/jlens.py`, `topogpt3/model.py`, `topogpt3/train.py`, `topogpt3/vanilla_control.py`
- `standard` | files=10 | mentions=14 | `app.py`, `eval/harness.py`, `eval/report.py`, `eval/samplers.py`, `eval/sandbox.py`, `eval/smoke.py`, `tests/test_jlens.py`, `topogpt3/__init__.py`, `topogpt3/continuation.py`, `topogpt3/inference_hrm.py`
- `python` | files=10 | mentions=13 | `eval/analyze_results.py`, `eval/governor.py`, `eval/harness.py`, `eval/noise_analysis.py`, `eval/sandbox.py`, `synthetic_dataset.py`, `topogpt3/api_server.py`, `topogpt3/jlens.py`, `topogpt3/model.py`, `transfer_weights.py`
- `tokens` | files=9 | mentions=64 | `eval/governor.py`, `eval/noise_analysis.py`, `topogpt3/__init__.py`, `topogpt3/inference.py`, `topogpt3/inference_hrm.py`, `topogpt3/jlens.py`, `topogpt3/model.py`, `topogpt3/train.py`, `topogpt3/vanilla_control.py`

## Verb Edges

- `topo` --depends_on--> `topogpt3` (strength 1.00)
- `topo` --depends_on--> `model` (strength 0.96)
- `gpt3` --depends_on--> `topogpt3` (strength 0.85)
- `gpt3` --depends_on--> `model` (strength 0.81)
- `gpt3` --depends_on--> `topo` (strength 0.81)
- `topo` --depends_on--> `config` (strength 0.79)
- `topo` --depends_on--> `new` (strength 0.79)
- `model` --depends_on--> `topogpt3` (strength 0.77)
- `topo` --depends_on--> `checkpoint` (strength 0.75)
- `topo` --depends_on--> `returns` (strength 0.75)
- `topo` --depends_on--> `file` (strength 0.73)
- `model` --depends_on--> `topo` (strength 0.71)
- `topo` --depends_on--> `all` (strength 0.71)
- `topo` --depends_on--> `build` (strength 0.71)
- `topo` --depends_on--> `prompt` (strength 0.71)
- `topo` --depends_on--> `run` (strength 0.71)
- `checkpoint` --depends_on--> `topogpt3` (strength 0.69)
- `gpt3` --depends_on--> `new` (strength 0.69)
- `topo` --depends_on--> `when` (strength 0.69)
- `checkpoint` --depends_on--> `model` (strength 0.67)
- `gpt3` --depends_on--> `config` (strength 0.67)
- `topo` --depends_on--> `gpt2` (strength 0.67)
- `topo` --depends_on--> `layer` (strength 0.67)
- `topo` --depends_on--> `load` (strength 0.67)
- `checkpoint` --depends_on--> `topo` (strength 0.65)
- `gpt3` --depends_on--> `checkpoint` (strength 0.65)
- `topo` --depends_on--> `max` (strength 0.65)
- `topo` --depends_on--> `tensor` (strength 0.65)
- `topogpt3` --depends_on--> `model` (strength 0.65)
- `gpt3` --depends_on--> `returns` (strength 0.62)
- `model` --depends_on--> `returns` (strength 0.62)
- `topo` --depends_on--> `shape` (strength 0.62)
- `topo` --depends_on--> `token` (strength 0.62)
- `topo` --depends_on--> `tokens` (strength 0.62)
- `topogpt3` --depends_on--> `topo` (strength 0.62)
- `gpt3` --depends_on--> `prompt` (strength 0.60)
- `model` --depends_on--> `new` (strength 0.60)
- `topo` --depends_on--> `data` (strength 0.60)
- `topo` --depends_on--> `per` (strength 0.60)
- `topo` --depends_on--> `true` (strength 0.60)
- `gpt3` --depends_on--> `all` (strength 0.58)
- `gpt3` --depends_on--> `build` (strength 0.58)
- `gpt3` --depends_on--> `file` (strength 0.58)
- `gpt3` --depends_on--> `run` (strength 0.58)
- `gpt3` --depends_on--> `when` (strength 0.58)
- `model` --depends_on--> `config` (strength 0.58)
- `topo` --depends_on--> `gpt3` (strength 0.58)
- `topo` --depends_on--> `loader` (strength 0.58)
- `topo` --depends_on--> `training` (strength 0.58)
- `gpt3` --depends_on--> `layer` (strength 0.56)

## Dialectic

- Thesis: `all` centralizes 14 files; Antithesis: `checkpoint` pulls 15 files with 9 shared (Jaccard 0.45); Synthesis: should they merge, split by layer, or keep `bridges` explicit?
- Thesis: `all` centralizes 14 files; Antithesis: `config` pulls 15 files with 9 shared (Jaccard 0.45); Synthesis: should they merge, split by layer, or keep `bridges` explicit?
- Thesis: `all` centralizes 14 files; Antithesis: `each` pulls 11 files with 6 shared (Jaccard 0.32); Synthesis: should they merge, split by layer, or keep `bridges` explicit?
- Thesis: `all` centralizes 14 files; Antithesis: `error` pulls 12 files with 7 shared (Jaccard 0.37); Synthesis: should they merge, split by layer, or keep `bridges` explicit?
- Thesis: `all` centralizes 14 files; Antithesis: `file` pulls 12 files with 10 shared (Jaccard 0.62); Synthesis: should they merge, split by layer, or keep `bridges` explicit?
- Thesis: `all` centralizes 14 files; Antithesis: `full` pulls 10 files with 6 shared (Jaccard 0.33); Synthesis: should they merge, split by layer, or keep `bridges` explicit?
- Thesis: `all` centralizes 14 files; Antithesis: `gpt2` pulls 10 files with 6 shared (Jaccard 0.33); Synthesis: should they merge, split by layer, or keep `bridges` explicit?
- Thesis: `all` centralizes 14 files; Antithesis: `gpt3` pulls 20 files with 9 shared (Jaccard 0.36); Synthesis: should they merge, split by layer, or keep `depends_on` explicit?
- Thesis: `all` centralizes 14 files; Antithesis: `layer` pulls 13 files with 9 shared (Jaccard 0.50); Synthesis: should they merge, split by layer, or keep `bridges` explicit?
- Thesis: `all` centralizes 14 files; Antithesis: `load` pulls 19 files with 8 shared (Jaccard 0.32); Synthesis: should they merge, split by layer, or keep `bridges` explicit?