simple_qwen3.5-397b_reactive_raw
Qwen3.5-397B on the simple floor, 5 episodes. Not cited by the thesis (development or earlier run).
| Wins (boss defeated, alive) | 0/5 [0.00, 0.43] |
| Final HP / exact optimum, mean over wins | – |
| Died | 0 |
| Final HP, mean | 469 |
| Seconds per model call | 325.9 |
| Parse error rate | 192.6% |
| Floor | Simple floor |
| Model | Qwen3.5-397B (qwen3.5-397b-a17b, thinking) |
| Planning | reactive |
| Prompt | engine-default-seed |
| Temperature | 0 |
| Lookahead depth | 3 |
| Mechanical overrides | off (all engine replacements and assists disabled) |
| # | outcome | final HP | HP / opt | steps | model calls | replay |
|---|---|---|---|---|---|---|
| 1 | not won | 474 | – | 36 | 17 | log not stored |
| 2 | not won | 541 | – | 36 | 16 | log not stored |
| 3 | not won | 443 | – | 36 | 16 | log not stored |
| 4 | not won | 443 | – | 36 | 16 | log not stored |
| 5 | not won | 443 | – | 36 | 16 | log not stored |
In a checkout of mt-ai-lab with an API key set (the result goes to a new file rerun_<time>_simple_qwen3.5-397b_reactive_raw, never over the thesis result):
python scripts/run_catalog.py --codes A2
Result file: results/reliability/simple_qwen3.5-397b_reactive_raw.json · updated 2026-09-23