AI Paper Atlas

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

ProbeLLM: Automating Principled Diagnosis of LLM Failures

Yue Huang, Zhengzhe Jiang, Yuchen Ma, Yu Jiang, Xiangqi Wang, Yujun Zhou, Yuexing Hao, Kehan Guo, Pin-Yu Chen, Marzyeh Ghassemi, Stefan Feuerriegel, Xiangliang Zhang

ICML 2026 · Large language models › Agents, code and math · approach: Code generation benchmarks · 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 ProbeLLM hierarchical MCTS probing instead of AutoDetectLlama-3.1-8b-ins. target · MMLUerror rate (%): 67% -> 86%Improved
use ProbeLLM instead of PAIRLlama-3.1-8b-ins. target · MMLUerror rate (%): 46% -> 86%Improved
use ProbeLLM instead of AutoDetect on SuperGLUE seedsLlama-3.1-8b-ins. target · SuperGLUEerror rate (%): 38% -> 71%Improved
use ProbeLLM instead of PAIR or AutoDetect for clustering failuresLlama-3.1-8b-ins. target · failure clustersnoise rate (%): 60.9 -> 37.0 (AutoDetect), 77.2 -> 37.0 (PAIR)Improved
probe with ProbeLLM instead of static benchmarksLlama-3.1-8b-ins. target · failure mode discoverynumber of clusters: 4 -> 19Improved
increase MCTS search depth (ablation)ProbeLLM, 12 target models · MMLU, MBPP, HellaSwag, TruthfulQAerror rate (%)Improved
increase MCTS search depth to 4 on SuperGLUE (ablation)ProbeLLM · SuperGLUEerror rate (%)Hurt
use tool-augmented generation (Python exec, web search) instead of perturbation only (ablation)ProbeLLM · tool usage analysiserror rate (%): 45.0% -> 60.7%Improved
vary MACRO/MICRO simulation ratio (beta) (ablation)ProbeLLM · five seed benchmarkserror rate (%)No clear effect
fine-tune on ProbeLLM-discovered failure data (100 to 250 samples)Llama-3.1-8b-ins. and GPT-oss-20b · error-prone evaluation setperformance (%)Mixed
fine-tune on ProbeLLM-discovered failure dataLlama-3.1-8b-ins. and GPT-oss-20b · held-out existing benchmark subsetsperformance (%)No clear effect
vary HDBSCAN clustering hyperparameters (ablation)ProbeLLM failure modes · clustering stabilitycluster stabilityNo clear effect

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