Cell · cited by the thesis

hand01_deepseek-v4-flash_final2_plan5

DeepSeek-V4-Flash on hand01, 5 episodes. Cited by the thesis in:

Outcome
Wins (boss defeated, alive)0/5 [0.00, 0.43]
Final HP / exact optimum, mean over wins–
Died0
Final HP, mean143
Seconds per model call8.7
Parse error rate0.0%
Settings
Floorhand01
ModelDeepSeek-V4-Flash (deepseek-v4-flash-0731, thinking)
Planningreactive
Promptengine-default-floor-basic
Temperature0
Lookahead depth3
Mechanical overrideson (logged per turn)
Engine settings (environment, 6)
THINKING_MAX_TOKENS=32768
CONVERSATION_HISTORY_TURNS=0
MOTA_ORDER_TWO_SIDED=1
MOTA_FIGHT_KEY_FLOOD=1
MOTA_PLAN_STEPS=5
LLM_TIMEOUT=1800
Episodesopen one to replay it turn by turn
#outcomefinal HPHP / optstepsmodel callsreplay
1not won145–326replay
2not won128–326replay
3not won153–328replay
4not won145–326replay
5not won145–326replay
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_deepseek-v4-flash_final2_plan5, never over the thesis result):

python scripts/run_catalog.py --codes HD1P5

Result file: results/reliability/hand01_deepseek-v4-flash_final2_plan5.json · updated 2026-09-30