Cell · cited by the thesis

hand04_deepseek-v4-flash_final2_t06

DeepSeek-V4-Flash on hand04, 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, mean114
Seconds per model call6.5
Parse error rate0.0%
Settings
Floorhand04
ModelDeepSeek-V4-Flash (deepseek-v4-flash-0731, thinking)
Planningreactive
Promptengine-default-floor-basic
Temperature0.6
Lookahead depth3
Mechanical overrideson (logged per turn)
Engine settings (environment, 5)
THINKING_MAX_TOKENS=16384
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 won108–3215replay
2not won107–3115replay
3not won107–3215replay
4not won114–3215replay
5not won135–3215replay
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_deepseek-v4-flash_final2_t06, never over the thesis result):

python scripts/run_catalog.py --codes T6HD4

Result file: results/reliability/hand04_deepseek-v4-flash_final2_t06.json · updated 2026-09-30