Agentic Abstention: Do Agents Know When to Stop Instead of Act?
cs.AI
Submitted: 2026-06-27
Updated: 2026-11-06
Comments: Accepted to NeurIPS 2026. Updated camera-ready version with revised references and minor corrections
Project page: https://lhannnn.github.io/agentic-abstention
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Terminology
Sources
- Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces
- Voyager: An Open-Ended Embodied Agent with Large Language Models
- Uncertainty Quantification in LLM Agents: Foundations, Emerging Challenges, and Opportunities
- Position: Uncertainty Quantification Needs Reassessment for Large-language Model Agents
- Check Yourself Before You Wreck Yourself: Selectively Quitting Improves LLM Agent Safety
- Active Task Disambiguation with LLMs
- Clarify or Answer: Reinforcement Learning for Agentic VQA with Context Under-specification
- AbstentionBench: Reasoning LLMs Fail on Unanswerable Questions
- AgentBench: Evaluating LLMs as Agents
- gpt-oss-120b & gpt-oss-20b Model Card
- On Evaluation of Embodied Navigation Agents
- Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models
- CValues: Measuring the Values of Chinese Large Language Models from Safety to Responsibility
- Latent Jailbreak: A Benchmark for Evaluating Text Safety and Output Robustness of Large Language Models
- All Languages Matter: On the Multilingual Safety of Large Language Models
- SORRY-Bench: Systematically Evaluating Large Language Model Safety Refusal
- Position: Agent Should Invoke External Tools ONLY When Epistemically Necessary
- GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning
- DialogGuard: Multi-Agent Psychosocial Safety Evaluation of Sensitive LLM Responses
- WorldSense: A Synthetic Benchmark for Grounded Reasoning in Large Language Models
Related papers
- Multi-Objective Structured Pruning of LLMs for Latency and Model Size Optimization
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- Embeddings for Preferences, Not Semantics
- Can We Trust Item Response Theory for AI Evaluation?
- Diagnosing Task Insensitivity in Language Agents