AI Paper Atlas

Research map / Large language models / Agents, code and math

ToC: Tree-of-Claims Search with Multi-Agent Language Models

Shuyang Yu, Jianan Liang, Hui Hu

AAAI 2026 · Large language models › Agents, code and math · approach: LLMs for optimization and heuristics · 1 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
add ToC search (MCTS with Examiner and Editor agents) over zero-shot promptingGPT-4o · USPTO wireless patent claimsoverall composite score: 0.647 -> 0.701Improved
add ToC search over few-shot promptingGPT-4o · USPTO wireless patent claimsoverall composite score: 0.678 -> 0.701Improved
add ToC search over zero-shot promptingOpenAI O1 · USPTO wireless patent claimsoverall composite score: 0.631 -> 0.680Improved
add ToC search over zero-shot promptingQwen2.5-VL-32B · USPTO wireless patent claimsoverall composite score: 0.592 -> 0.639Improved
add ToC search over zero-shot promptingQwen2.5-VL-72B · USPTO wireless patent claimsoverall composite score: 0.611 -> 0.658Improved
add ToC search over zero-shot promptingGPT-4o · USPTO wireless patent claimscoverage reward: 0.520 -> 0.582Improved
add ToC search over zero-shot promptingGPT-4o · USPTO wireless patent claimsscope reward (penalty term): 0.417 -> 0.389Improved
add ToC search over zero-shot promptingGPT-4o · USPTO wireless patent claimsperplexity: 8.93 -> 8.72Improved
add ToC search over zero-shot promptingGPT-4o · USPTO wireless patent claimsJSON parse completeness: 0.994 -> 0.996No clear effect
use few-shot prompting with chain-of-thought exemplars instead of zero-shotGPT-4o · USPTO wireless patent claimsoverall composite score: 0.647 -> 0.678Improved
disable uncertainty gating, progressive widening, or multi-agent collaboration (ablation)ToC · USPTO wireless patent claimsnovelty, coverage, legal qualityHurt
use progressive widening alpha 0.8 or search depth at most 5 (ablation)ToC sensitivity · USPTO wireless patent claimsaccuracy: 3-5 points lowerHurt

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