Cell · cited by the thesis

complex_deepseek-v4-flash_canon_hp60

DeepSeek-V4-Flash on the complex floor, 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, mean116
Seconds per model call16.7
Parse error rate0.0%
Settings
FloorComplex floor
ModelDeepSeek-V4-Flash (deepseek-v4-flash-0731, thinking)
Planningreactive
Promptfloor_basic_hp60.txt
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_HP_RULE_PCT=60
LLM_TIMEOUT=1800
Episodesopen one to replay it turn by turn
#outcomefinal HPHP / optstepsmodel callsreplay
1not won200–2611replay
2not won123–3715replay
3not won98–3715replay
4not won25–3715replay
5not won134–3514replay
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>_complex_deepseek-v4-flash_canon_hp60, never over the thesis result):

python scripts/run_catalog.py --codes H6DC

Result file: results/reliability/complex_deepseek-v4-flash_canon_hp60.json · updated 2026-10-03