Papers13

#calibration

Towards Reducible Uncertainty Modeling for Reliable Large Language Model Agents

Changdae Oh, Seongheon Park et al.Feb 4arXiv

This paper says we should measure an AI agent’s uncertainty across its whole conversation, not just on one final answer.

#uncertainty quantification#LLM agents#interactive AI

Accurate Failure Prediction in Agents Does Not Imply Effective Failure Prevention

Intermediate

Rakshith Vasudev, Melisa Russak et al.Feb 3arXiv

The paper shows that even if a model is great at predicting when an AI agent will fail, jumping in to “fix” the agent mid-task can still make things worse.

#LLM critic#execution-time intervention#disruption–recovery tradeoff

Rethinking Selective Knowledge Distillation

Intermediate

Almog Tavor, Itay Ebenspanger et al.Feb 1arXiv

The paper studies how to teach a smaller language model using a bigger one by only focusing on the most useful bits instead of everything.

#knowledge distillation#selective distillation#student entropy

Deep Search with Hierarchical Meta-Cognitive Monitoring Inspired by Cognitive Neuroscience

Intermediate

Zhongxiang Sun, Qipeng Wang et al.Jan 30arXiv

Deep search agents can plan and browse the web in many steps, but they often fail because they don’t notice when their own thinking drifts off-track.

#deep search agents#metacognition#consistency monitoring

WorldVQA: Measuring Atomic World Knowledge in Multimodal Large Language Models

Intermediate

Runjie Zhou, Youbo Shao et al.Jan 28arXiv

WorldVQA is a new test that checks if multimodal AI models can correctly name what they see in pictures without doing extra reasoning.

#WorldVQA#atomic visual knowledge#multimodal large language models

PROGRESSLM: Towards Progress Reasoning in Vision-Language Models

Intermediate

Jianshu Zhang, Chengxuan Qian et al.Jan 21arXiv

This paper asks a new question for vision-language models: not just 'What do you see?' but 'How far along is the task right now?'

#progress reasoning#vision-language models#episodic retrieval

Entropy Sentinel: Continuous LLM Accuracy Monitoring from Decoding Entropy Traces in STEM

Beginner

Pedro Memoli Buffa, Luciano Del CorroJan 13arXiv

The paper introduces Entropy Sentinel, a simple way to watch how accurate an AI is by reading its “uncertainty heartbeat” during generation.

#LLM monitoring#entropy profile#top-k probabilities

The Confidence Dichotomy: Analyzing and Mitigating Miscalibration in Tool-Use Agents

Beginner

Weihao Xuan, Qingcheng Zeng et al.Jan 12arXiv

This paper studies how AI agents that use tools talk about how sure they are and finds a split: some tools make them too sure, others help them be honest.

#LLM agents#calibration#overconfidence

EpiCaR: Knowing What You Don't Know Matters for Better Reasoning in LLMs

Intermediate

Jewon Yeom, Jaewon Sok et al.Jan 11arXiv

This paper teaches AI models not just how to solve problems but also how to tell when their own answers might be wrong.

#EPICAR#calibration#epistemic uncertainty

SWE-RM: Execution-free Feedback For Software Engineering Agents

Intermediate

KaShun Shum, Binyuan Hui et al.Dec 26arXiv

Coding agents used to fix software rely on feedback; unit tests give only pass/fail signals that are often noisy or missing.

#execution-free feedback#reward model#software engineering agents

Can LLMs Predict Their Own Failures? Self-Awareness via Internal Circuits

Intermediate

Amirhosein Ghasemabadi, Di NiuDec 23arXiv

Large language models often sound confident even when they are wrong, and existing ways to catch mistakes are slow or not very accurate.

#self-awareness#large language models#hidden states

Causal Judge Evaluation: Calibrated Surrogate Metrics for LLM Systems

Intermediate

Eddie Landesberg, Manjari NarayanDec 11arXiv

LLM judges are cheap but biased; without calibration they can completely flip which model looks best.

#LLM-as-judge#calibration#isotonic regression

1 2