Everything behind the paper: the data, the code that produced every table, and the ledger of every claim considered, including the ones that failed checking.
| file | size | what it is |
|---|---|---|
| PLAN.md | 18.0 KB | a note published with the paper |
| analyse.py | 15.8 KB | the code that produced or checked the numbers |
| claim-ledger.csv | 20.0 KB | every claim in the paper, its status and its source |
| controls.py | 4.0 KB | the code that produced or checked the numbers |
| dsh_adapter.py | 11.3 KB | the code that produced or checked the numbers |
| floor.json | 0.4 KB | structured output kept exactly as recorded |
| graders08.py | 8.1 KB | the code that produced or checked the numbers |
| how-to-run.md | 3.6 KB | a note published with the paper |
| integrity.py | 4.8 KB | recomputes every headline figure and fails if the paper no longer says it |
| openclaw_adapter.py | 14.1 KB | the code that produced or checked the numbers |
| source-log.csv | 4.3 KB | every source consulted, with its retrieval date |
| suiteW.py | 6.6 KB | the code that produced or checked the numbers |
| sweep.py | 26.4 KB | the code that produced or checked the numbers |
| tables.py | 16.9 KB | the code that produced or checked the numbers |
| tiers.py | 22.3 KB | the code that produced or checked the numbers |
| winagent_adapter.py | 9.7 KB | the code that produced or checked the numbers |
| sessions/axium__S-high__V.jsonl | 73.5 KB | one graded session, every turn, tool call, token and verdict |
| sessions/axium__S-high__W.jsonl | 39.7 KB | one graded session, every turn, tool call, token and verdict |
| sessions/axium__S-off__V.jsonl | 71.9 KB | one graded session, every turn, tool call, token and verdict |
| sessions/axium__S-off__W.jsonl | 40.0 KB | one graded session, every turn, tool call, token and verdict |
| sessions/axium__T0__V.jsonl | 37.7 KB | one graded session, every turn, tool call, token and verdict |
| sessions/axium__T0__W.jsonl | 21.8 KB | one graded session, every turn, tool call, token and verdict |
| sessions/axium__T1__V.jsonl | 75.9 KB | one graded session, every turn, tool call, token and verdict |
| sessions/axium__T1__W.jsonl | 38.7 KB | one graded session, every turn, tool call, token and verdict |
| sessions/axium__T2__V.jsonl | 68.7 KB | one graded session, every turn, tool call, token and verdict |
| sessions/axium__T2__W.jsonl | 35.1 KB | one graded session, every turn, tool call, token and verdict |
| sessions/axium__T3__V.jsonl | 57.7 KB | one graded session, every turn, tool call, token and verdict |
| sessions/axium__T3__W.jsonl | 31.7 KB | one graded session, every turn, tool call, token and verdict |
| sessions/dsh__S-high__V.jsonl | 78.1 KB | one graded session, every turn, tool call, token and verdict |
| sessions/dsh__S-high__W.jsonl | 34.5 KB | one graded session, every turn, tool call, token and verdict |
| sessions/dsh__S-off__W.jsonl | 20.3 KB | one graded session, every turn, tool call, token and verdict |
| sessions/dsh__T0__V.jsonl | 57.2 KB | one graded session, every turn, tool call, token and verdict |
| sessions/dsh__T0__W.jsonl | 27.2 KB | one graded session, every turn, tool call, token and verdict |
| sessions/dsh__T1__V.jsonl | 82.6 KB | one graded session, every turn, tool call, token and verdict |
| sessions/dsh__T1__W.jsonl | 34.9 KB | one graded session, every turn, tool call, token and verdict |
| sessions/dsh__T2__V.jsonl | 71.8 KB | one graded session, every turn, tool call, token and verdict |
| sessions/dsh__T2__W.jsonl | 33.1 KB | one graded session, every turn, tool call, token and verdict |
| sessions/dsh__T3__V.jsonl | 53.1 KB | one graded session, every turn, tool call, token and verdict |
| sessions/dsh__T3__W.jsonl | 27.2 KB | one graded session, every turn, tool call, token and verdict |
| sessions/hermes__T0__V.jsonl | 49.6 KB | one graded session, every turn, tool call, token and verdict |
| sessions/hermes__T0__W.jsonl | 24.5 KB | one graded session, every turn, tool call, token and verdict |
| sessions/hermes__T1__V.jsonl | 78.8 KB | one graded session, every turn, tool call, token and verdict |
| sessions/hermes__T1__W.jsonl | 39.5 KB | one graded session, every turn, tool call, token and verdict |
| sessions/hermes__T2__V.jsonl | 76.0 KB | one graded session, every turn, tool call, token and verdict |
| sessions/hermes__T2__W.jsonl | 40.7 KB | one graded session, every turn, tool call, token and verdict |
| sessions/hermes__T3__V.jsonl | 60.5 KB | one graded session, every turn, tool call, token and verdict |
| sessions/hermes__T3__W.jsonl | 33.1 KB | one graded session, every turn, tool call, token and verdict |
| sessions/null__none__V.jsonl | 33.0 KB | one graded session, every turn, tool call, token and verdict |
| sessions/null__none__W.jsonl | 7.6 KB | one graded session, every turn, tool call, token and verdict |
| sessions/openclaw__S-high__V.jsonl | 70.1 KB | one graded session, every turn, tool call, token and verdict |
| sessions/openclaw__S-high__W.jsonl | 32.2 KB | one graded session, every turn, tool call, token and verdict |
| sessions/openclaw__S-off__V.jsonl | 69.6 KB | one graded session, every turn, tool call, token and verdict |
| sessions/openclaw__S-off__W.jsonl | 33.9 KB | one graded session, every turn, tool call, token and verdict |
| sessions/openclaw__T0__V.jsonl | 60.7 KB | one graded session, every turn, tool call, token and verdict |
| sessions/openclaw__T0__W.jsonl | 27.0 KB | one graded session, every turn, tool call, token and verdict |
| sessions/openclaw__T1__V.jsonl | 69.6 KB | one graded session, every turn, tool call, token and verdict |
| sessions/openclaw__T1__W.jsonl | 31.7 KB | one graded session, every turn, tool call, token and verdict |
| sessions/openclaw__T2__V.jsonl | 63.1 KB | one graded session, every turn, tool call, token and verdict |
| sessions/openclaw__T2__W.jsonl | 30.1 KB | one graded session, every turn, tool call, token and verdict |
| sessions/openclaw__T3__V.jsonl | 50.9 KB | one graded session, every turn, tool call, token and verdict |
| sessions/openclaw__T3__W.jsonl | 26.5 KB | one graded session, every turn, tool call, token and verdict |
| sessions/orange__S-high__V.jsonl | 55.8 KB | one graded session, every turn, tool call, token and verdict |
| sessions/orange__S-high__W.jsonl | 25.6 KB | one graded session, every turn, tool call, token and verdict |
| sessions/orange__S-off__V.jsonl | 55.6 KB | one graded session, every turn, tool call, token and verdict |
| sessions/orange__S-off__W.jsonl | 27.9 KB | one graded session, every turn, tool call, token and verdict |
| sessions/orange__T0__V.jsonl | 45.5 KB | one graded session, every turn, tool call, token and verdict |
| sessions/orange__T0__W.jsonl | 22.9 KB | one graded session, every turn, tool call, token and verdict |
| sessions/orange__T1__V.jsonl | 52.0 KB | one graded session, every turn, tool call, token and verdict |
| sessions/orange__T1__W.jsonl | 25.3 KB | one graded session, every turn, tool call, token and verdict |
| sessions/orange__T2__V.jsonl | 54.0 KB | one graded session, every turn, tool call, token and verdict |
| sessions/orange__T2__W.jsonl | 25.1 KB | one graded session, every turn, tool call, token and verdict |
| sessions/orange__T3__V.jsonl | 50.0 KB | one graded session, every turn, tool call, token and verdict |
| sessions/orange__T3__W.jsonl | 25.5 KB | one graded session, every turn, tool call, token and verdict |
| sessions/prose__none__V.jsonl | 54.6 KB | one graded session, every turn, tool call, token and verdict |
| sessions/prose__none__W.jsonl | 9.9 KB | one graded session, every turn, tool call, token and verdict |
| sessions/winagent__T0__V.jsonl | 40.9 KB | one graded session, every turn, tool call, token and verdict |
| sessions/winagent__T0__W.jsonl | 27.0 KB | one graded session, every turn, tool call, token and verdict |
| sessions/winagent__T1__V.jsonl | 80.9 KB | one graded session, every turn, tool call, token and verdict |
| sessions/winagent__T1__W.jsonl | 36.6 KB | one graded session, every turn, tool call, token and verdict |
| sessions/winagent__T2__V.jsonl | 69.3 KB | one graded session, every turn, tool call, token and verdict |
| sessions/winagent__T2__W.jsonl | 34.1 KB | one graded session, every turn, tool call, token and verdict |
| sessions/winagent__T3__V.jsonl | 105.6 KB | one graded session, every turn, tool call, token and verdict |
| sessions/winagent__T3__W.jsonl | 47.0 KB | one graded session, every turn, tool call, token and verdict |
| sessions/invalid/README.md | 2.9 KB | how to run the code in this package |
| sessions/invalid/axium__T0__W.insufficient-balance.jsonl | 0.0 KB | one record per line, as the run produced it |
| sessions/invalid/axium__T1__V.insufficient-balance.jsonl | 18.1 KB | one record per line, as the run produced it |
| sessions/invalid/axium__T1__W.insufficient-balance.jsonl | 11.2 KB | one record per line, as the run produced it |
| sessions/invalid/axium__T2__V.insufficient-balance.jsonl | 18.1 KB | one record per line, as the run produced it |
| sessions/invalid/axium__T2__W.insufficient-balance.jsonl | 11.2 KB | one record per line, as the run produced it |
| sessions/invalid/dsh__S-off__V.cannot-express-off.jsonl | 28.8 KB | one record per line, as the run produced it |
| sessions/invalid/dsh__T1__V.boot-failure.jsonl | 13.3 KB | one record per line, as the run produced it |
| sessions/invalid/dsh__T1__V.insufficient-balance.jsonl | 17.1 KB | one record per line, as the run produced it |
| sessions/invalid/dsh__T1__W.insufficient-balance.jsonl | 10.7 KB | one record per line, as the run produced it |
| sessions/invalid/dsh__T2__V.insufficient-balance.jsonl | 17.1 KB | one record per line, as the run produced it |
| sessions/invalid/dsh__T2__W.insufficient-balance.jsonl | 10.6 KB | one record per line, as the run produced it |
| sessions/invalid/hermes__T0__V.tools-never-fired.jsonl | 0.0 KB | one record per line, as the run produced it |
| sessions/invalid/hermes__T1__V.insufficient-balance.jsonl | 20.3 KB | one record per line, as the run produced it |
| sessions/invalid/hermes__T1__V.tools-never-fired.jsonl | 0.0 KB | one record per line, as the run produced it |
| sessions/invalid/hermes__T1__W.insufficient-balance.jsonl | 11.8 KB | one record per line, as the run produced it |
| sessions/invalid/hermes__T2__V.insufficient-balance.jsonl | 20.2 KB | one record per line, as the run produced it |
| sessions/invalid/hermes__T2__W.insufficient-balance.jsonl | 11.7 KB | one record per line, as the run produced it |
| sessions/invalid/openclaw__T1__V.insufficient-balance.jsonl | 24.6 KB | one record per line, as the run produced it |
| sessions/invalid/openclaw__T1__W.insufficient-balance.jsonl | 13.2 KB | one record per line, as the run produced it |
| sessions/invalid/openclaw__T2__V.insufficient-balance.jsonl | 24.5 KB | one record per line, as the run produced it |
| sessions/invalid/openclaw__T2__W.insufficient-balance.jsonl | 13.1 KB | one record per line, as the run produced it |
| sessions/invalid/orange__T1__V.insufficient-balance.jsonl | 14.4 KB | one record per line, as the run produced it |
| sessions/invalid/orange__T1__W.insufficient-balance.jsonl | 9.8 KB | one record per line, as the run produced it |
| sessions/invalid/orange__T2__V.insufficient-balance.jsonl | 14.3 KB | one record per line, as the run produced it |
| sessions/invalid/orange__T2__W.insufficient-balance.jsonl | 9.8 KB | one record per line, as the run produced it |
| sessions/invalid/winagent__T1__V.insufficient-balance.jsonl | 16.2 KB | one record per line, as the run produced it |
| sessions/invalid/winagent__T1__W.insufficient-balance.jsonl | 10.6 KB | one record per line, as the run produced it |
| sessions/invalid/winagent__T2__V.insufficient-balance.jsonl | 16.2 KB | one record per line, as the run produced it |
| sessions/invalid/winagent__T2__W.insufficient-balance.jsonl | 10.5 KB | one record per line, as the run produced it |
| sessions/pre-correction/null__T0__V.jsonl | 26.9 KB | one graded session, every turn, tool call, token and verdict |
| sessions/pre-correction/prose__T0__V.jsonl | 48.5 KB | one graded session, every turn, tool call, token and verdict |
| sessions/superseded/README.md | 0.8 KB | how to run the code in this package |
| sessions/superseded/axium-guard__T0__V.pre-guard.jsonl | 12.1 KB | one record per line, as the run produced it |
| sessions/superseded/axium__T0__V.pre-guard.jsonl | 19.6 KB | one record per line, as the run produced it |
| sessions/superseded/axium__T0__W.pre-guard.jsonl | 10.6 KB | one record per line, as the run produced it |
| sessions/superseded/axium__T1__V.pre-guard.jsonl | 35.8 KB | one record per line, as the run produced it |
| sessions/superseded/axium__T2__V.pre-guard.jsonl | 31.0 KB | one record per line, as the run produced it |
Every figure in the paper can be recomputed from these files. If one of them disagrees with the paper, the file is right and the paper is wrong, and it will be corrected in the change log.