Cell

hand04_qwen3.5-397b_final2

Qwen3.5-397B on hand04, 5 episodes. Not cited by the thesis (development or earlier run).

Outcome
Wins (boss defeated, alive)0/5 [0.00, 0.43]
Final HP / exact optimum, mean over wins–
Died0
Final HP, mean113
Seconds per model call73.7
Parse error rate0.0%
Settings
Floorhand04
ModelQwen3.5-397B (qwen3.5-397b-a17b, thinking)
Planningreactive
Promptengine-default-floor-basic
Temperature0
Lookahead depth3
Mechanical overrideson (logged per turn)
Engine settings (environment, 5)
THINKING_MAX_TOKENS=32768
CONVERSATION_HISTORY_TURNS=0
MOTA_ORDER_TWO_SIDED=1
MOTA_FIGHT_KEY_FLOOD=1
LLM_TIMEOUT=300
Episodesopen one to replay it turn by turn
#outcomefinal HPHP / optstepsmodel callsreplay
1not won119–3215replay
2not won110–3115replay
3not won114–3115replay
4not won104–3115replay
5not won118–3215replay
Re-run this cell

In a checkout of mt-ai-lab with an API key set (the result goes to a new file rerun_<time>_hand04_qwen3.5-397b_final2, never over the thesis result):

python scripts/run_catalog.py --codes HQ4

Result file: results/reliability/hand04_qwen3.5-397b_final2.json · updated 2026-09-30