Cell · cited by the thesis

hand01_mistral-medium-3.5_canon

Mistral-Medium-3.5 on hand01, 5 episodes. Cited by the thesis in:

Outcome
Wins (boss defeated, alive)5/5 [0.57, 1.00]
Final HP / exact optimum, mean over wins64%
Died0
Final HP, mean9
Seconds per model call32.5
Parse error rate0.0%
Settings
Floorhand01
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
1won964%2813replay
2won964%2813replay
3won964%2813replay
4won964%2813replay
5won964%2813replay
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>_hand01_mistral-medium-3.5_canon, never over the thesis result):

python scripts/run_catalog.py --codes VM1 VM1a VM15

Result file: results/reliability/hand01_mistral-medium-3.5_canon.json · updated 2026-10-03