AI Paper Atlas

Research map / Multi-agent systems and games / Human-AI and goal-conditioned

MINT: Minimal Information Neuro-Symbolic Tree for Objective-Driven Knowledge-Gap Reasoning and Active Elicitation

Zeyu Fang, Mahdi Imani, Tian Lan

ICML 2026 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Human-AI coordination agents · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use MINT query elicitation instead of pure-LLM planningMINT vs o3 / GPT-4o · MiniGrid (5 uncertain objects)average reward: 7.81 -> 9.69Improved
use MINT instead of pure RL (DQN) under many unknown objectsDQN · MiniGrid (5 uncertain objects)success rate (%): 15.2 -> 98.2Improved
cap number of queries per unknown object (MINT Limited vs Standard) (ablation)MINT · MiniGrid (5 uncertain objects)average reward: 9.69 -> 9.56No clear effect
cap number of queries (Limited vs Standard) (ablation)MINT · Atari Pacmanreturn: 434.9 -> 411.1 return; 6.8 -> 3.8 queriesMixed
replace expert-action queries (Query-A) with MINT binary queriesHITL-RL baseline · Atari Pacmanquery times: 27.1 -> 6.8Improved
add MINT elicitation on top of LLM planning with fixed treeLLM planner · NVIDIA Isaac hidden target (target 2)success rate (%): 9 -> 95 (o3 vs MINT*)Improved
add uncertain objects (0 to 5) for pure RL DQNDQN · MiniGridaverage reward: 9.32 -> -2.01Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy