Cell · cited by the thesis

hand02_deepseek-v4-flash_final2_reactive_t32

DeepSeek-V4-Flash on hand02, 5 episodes. Cited by the thesis in:

Outcome
Wins (boss defeated, alive)4/5 [0.38, 0.96]
Final HP / exact optimum, mean over wins29%
Died0
Final HP, mean15
Seconds per model call7.6
Parse error rate0.0%
Settings
Floorhand02
ModelDeepSeek-V4-Flash (deepseek-v4-flash-0731, 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=1800
Episodesopen one to replay it turn by turn
#outcomefinal HPHP / optstepsmodel callsreplay
1won431%279replay
2won323%279replay
3won431%279replay
4won431%279replay
5not won62–3113replay
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_deepseek-v4-flash_final2_reactive_t32, never over the thesis result):

python scripts/run_catalog.py --codes HD2R32

Result file: results/reliability/hand02_deepseek-v4-flash_final2_reactive_t32.json · updated 2026-09-30