Cell · cited by the thesis

hand02_mistral-medium-3.5_canon

Mistral-Medium-3.5 on hand02, 5 episodes. Cited by the thesis in:

Outcome
Wins (boss defeated, alive)5/5 [0.57, 1.00]
Final HP / exact optimum, mean over wins86%
Died0
Final HP, mean11
Seconds per model call17.9
Parse error rate2.2%
Settings
Floorhand02
ModelMistral-Medium-3.5 (mistral-medium-3.5-128b, instruct)
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
1won431%279replay
2won13100%279replay
3won13100%279replay
4won13100%279replay
5won13100%279replay
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>_hand02_mistral-medium-3.5_canon, never over the thesis result):

python scripts/run_catalog.py --codes VM2 VM2a VM25

Result file: results/reliability/hand02_mistral-medium-3.5_canon.json · updated 2026-10-03