ECHO is a clipped policy-gradient method that uses posterior-sensitive rewards to give turn-level epistemic credit in multi-turn information-seeking tasks, outperforming trajectory-level GRPO on a new Clue Selector Game benchmark.
Active task disambiguation with llms.arXiv preprint arXiv:2502.04485
6 Pith papers cite this work. Polarity classification is still indexing.
6
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
verdicts
UNVERDICTED 6roles
background 1polarities
background 1representative citing papers
LLM agents often fail to abstain at the right time in uncertain multi-turn tasks, and the CONVOLVE context engineering method raises timely abstention rates on WebShop from 26.7 to 57.4 without parameter updates.
Ambig-IaC detects structural disagreements in LLM-generated IaC candidates across three hierarchical axes to produce clarification questions, improving structure and attribute accuracy by 18.4% and 25.4% on a new 300-task benchmark.
AI alignment must move beyond assuming users have fully formed goals and instead provide active cognitive support to help form and refine intent over time.
LILO integrates LLMs to translate natural language feedback into preference signals for Gaussian process-based Bayesian optimization, outperforming standard preference BO and LLM-only methods on benchmarks.
A tradeoff model shows generative AI can reduce bias against diverse preferences by strategically eliciting information instead of always inferring from majority patterns.
citing papers explorer
-
ECHO: Learning Epistemically Adaptive Language Agents with Turn-Level Credit
ECHO is a clipped policy-gradient method that uses posterior-sensitive rewards to give turn-level epistemic credit in multi-turn information-seeking tasks, outperforming trajectory-level GRPO on a new Clue Selector Game benchmark.
-
Agentic Abstention: Do Agents Know When to Stop Instead of Act?
LLM agents often fail to abstain at the right time in uncertain multi-turn tasks, and the CONVOLVE context engineering method raises timely abstention rates on WebShop from 26.7 to 57.4 without parameter updates.
-
Ambig-IaC: Multi-level Disambiguation for Interactive Cloud Infrastructure-as-Code Synthesis
Ambig-IaC detects structural disagreements in LLM-generated IaC candidates across three hierarchical axes to produce clarification questions, improving structure and attribute accuracy by 18.4% and 25.4% on a new 300-task benchmark.
-
Alignment has a Fantasia Problem
AI alignment must move beyond assuming users have fully formed goals and instead provide active cognitive support to help form and refine intent over time.
-
LILO: Bayesian Optimization with Natural Language Feedback
LILO integrates LLMs to translate natural language feedback into preference signals for Gaussian process-based Bayesian optimization, outperforming standard preference BO and LLM-only methods on benchmarks.
-
When to Ask a Question: Understanding Communication Strategies in Generative AI Tools
A tradeoff model shows generative AI can reduce bias against diverse preferences by strategically eliciting information instead of always inferring from majority patterns.