AI Paper Atlas

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

FormalJudge: A Neuro-Symbolic Paradigm for Agentic Oversight

Jiayi Zhou, Yang Sheng, Hantao Lou, Yaodong Yang, Jie Fu

ICML 2026 · Large language models › Agents, code and math · approach: Inductive and abstract reasoning evaluation · 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 FormalJudge instead of vanilla LLM judgeClaude-4.5-Opus judge, GPT-4o agent · Agent-SafetyBenchdetection accuracy (%): 58.00 -> 77.28 (+19.28)Improved
use FormalJudge instead of vanilla LLM judgeClaude-4.5-Opus judge, Gemini-3.0-Pro agent · Agent-SafetyBenchdetection accuracy (%): 58.00 -> 80.54 (+22.54)Improved
use FormalJudge instead of vanilla LLM judgeClaude-4.5-Opus judge, Qwen-32B agent · VitaBenchdetection accuracy (%): 66.05 -> 78.53 (+12.48)Improved
use FormalJudge instead of vanilla LLM judgeClaude-4.5-Opus judge, Qwen-32B agent · Deceiversdetection accuracy (%): 66.57 -> 96.92 (+30.35)Improved
add chain-of-thought prompting to the LLM judgeLLM judge · Agent-SafetyBenchdetection accuracy (%): 58.00 -> 61.50 (+3.50) on GPT-4o; 72.22 -> 69.70 (-2.52) on Claude-4.5-SonnetMixed
add chain-of-thought prompting to the LLM judgeLLM judge · VitaBenchdetection accuracy (%): 72.23 -> 65.84 (-6.39) on GPT-4o; 62.61 -> 68.42 (+5.81) on Claude-4.5-SonnetMixed
add structured trajectory information to the LLM judgeLLM judge · Agent-SafetyBenchdetection accuracy (%): 58.00 -> 76.60 (+18.60) on Gemini-3.0-ProImproved
add Python verification scripts to the LLM judgeLLM judge · Agent-SafetyBenchdetection accuracy (%): 58.00 -> 68.70 (+10.70) on GPT-4o; 72.22 -> 58.59 (-13.63) on Claude-4.5-SonnetMixed
add structured information to LLM judgeLLM judge · Deceiversdetection accuracy (%): 62.47 -> 71.58 (+9.11) on GPT-4oImproved
add Dafny formal verification feedback in iterative refinement (ablation)Claude-4.5-Opus agent · Agent-SafetyBench refinementsafety accuracy (%): 70.7% -> 99.8%; 29.1% absolute improvementImproved
use Dafny feedback instead of Few-shot or CoT feedback in refinement (ablation)Claude-4.5-Opus agent · Agent-SafetyBench refinement (round 3)safety accuracy (%): 83.5% and 85.5% -> 99.8%Improved
add Dafny feedback in refinement (ablation)Qwen-7B agent · Agent-SafetyBench refinementsafety accuracy (%): 30.8% -> 49.4%Improved

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