Cell

segment1_qwen3.5-397b_basic_d1d4

Qwen3.5-397B on the tower (MT5→MT8), 1 run. Not cited by the thesis (development or earlier run).

Outcome
Highest floor reached, per runMT8
Died0
Final HP, mean895
Seconds per model call28.2
Parse error rate4.9%
Settings
FloorTower MT5→MT8
ModelQwen3.5-397B (qwen3.5-397b-a17b, thinking)
Planningreactive
Promptengine-default-tower-basic
Temperature0
Lookahead depth2
Mechanical overrideson (logged per turn)
Walkthrough checkpoint1
Engine settings (environment, 1)
THINKING_MAX_TOKENS=8192
Runsopen one to replay it turn by turn
#outcomefinal HPHP / optturnsmodel callsreplay
1MT8895–9461replay
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>_segment1_qwen3.5-397b_basic_d1d4, never over the thesis result):

python scripts/run_catalog.py --codes W1

Result file: results/reliability/segment1_qwen3.5-397b_basic_d1d4.json · updated 2026-09-26