List your company

Text-based agent interaction on TextWorld Quest

88AccuracyAgent-BRACE
59.9267.2174.581.79May 12, 2026

Evaluation results

MethodLinks
Agent-BRACEBackbone=Qwen3-4B-Instruct, Independent runs=32026.058830.8
PABUBackbone=Qwen3-4B-Instruct, Independent runs=32026.058326.4
ReAct (RL)Backbone=Qwen3-4B-Instruct, Independent runs=32026.0575.518.3
Direct-Action (RL)Backbone=Qwen3-4B-Instruct, Independent runs=32026.057429.8
Base ModelBackbone=Qwen3-4B-Instruct, Independent runs=32026.0561.532.2
ReActBackbone=Qwen3-4B-Instruct, Independent runs=32026.056112.8