Cell

complex_glm-5.3-flash_final2_la3

GLM-5.3-Flash on the complex floor, 4 episodes. Not cited by the thesis (development or earlier run).

Outcome
Wins (boss defeated, alive)0/4 [0.00, 0.49]
Final HP / exact optimum, mean over wins–
Died0
Final HP, mean119
Seconds per model call905.3
Parse error rate52.5%
Settings
FloorComplex floor
ModelGLM-5.3-Flash (glm-5.3-flash, 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=300
Episodesopen one to replay it turn by turn
#outcomefinal HPHP / optstepsmodel callsreplay
1not won143–3915replay
2not won135–3715replay
3not won73–3916replay
4not won123–3715replay
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_glm-5.3-flash_final2_la3, never over the thesis result):

python scripts/run_catalog.py --codes GLC

Result file: results/reliability/complex_glm-5.3-flash_final2_la3.json · updated 2026-10-01