Cell

hand01_deepseek-v4-flash_blockergraph

DeepSeek-V4-Flash on hand01, 5 episodes. Not cited by the thesis (development or earlier run).

Outcome
Wins (boss defeated, alive)1/5 [0.04, 0.62]
Final HP / exact optimum, mean over wins71%
Died0
Final HP, mean62
Seconds per model call14.0
Parse error rate0.0%
Settings
Floorhand01
ModelDeepSeek-V4-Flash (deepseek-v4-flash-0731, thinking)
Planningreactive
Promptblocker_lookahead.txt
Temperature0
Lookahead depth3
Mechanical overrideson (logged per turn)
Engine settings (environment, 7)
THINKING_MAX_TOKENS=32768
CONVERSATION_HISTORY_TURNS=0
MOTA_ORDER_TWO_SIDED=1
MOTA_FIGHT_KEY_FLOOD=1
MOTA_BARE_OBS=1
MOTA_BLOCKER_GRAPH=1
LLM_TIMEOUT=1800
Episodesopen one to replay it turn by turn
#outcomefinal HPHP / optstepsmodel callsreplay
1not won48–179replay
2not won48–179replay
3won1071%2813replay
4not won154–3217replay
5not won48–179replay
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_blockergraph, never over the thesis result):

python scripts/run_catalog.py --codes BGH

Result file: results/reliability/hand01_deepseek-v4-flash_blockergraph.json · updated 2026-10-02