Useless but Safe? Benchmarking Utility Recovery with User Intent Clarification in Multi-Turn Conversations

• URL: http://arxiv.org/abs/2604.27093v1
• Date: Wed, 29 Apr 2026 18:37:18 GMT
• Title: Useless but Safe? Benchmarking Utility Recovery with User Intent Clarification in Multi-Turn Conversations
• Abstract summary: We introduce CarryOnBench, the first interactive benchmark that measures whether LLMs can revise their interpretation of user intent and recover utility.<n>We simulate 5,970 conversations by varying user follow-up sequences, evaluating 14 models on both intent-aligned utility and safety.<n> CarryOnBench yields 1,866 different conversation flows of 4--12 turns, totaling 23,880 model responses.
• Score: 32.23729177914094
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Current LLM safety alignment techniques improve model robustness against adversarial attacks, but overlook whether and how LLMs can recover helpfulness when benign users clarify their intent. We introduce CarryOnBench, the first interactive benchmark that measures whether LLMs can revise their interpretation of user intent and recover utility, while remaining safe through multi-turn conversations. Starting from 398 seemingly harmful queries with benign underlying intents, we simulate 5,970 conversations by varying user follow-up sequences, evaluating 14 models on both intent-aligned utility and safety. CarryOnBench yields 1,866 different conversation flows of 4--12 turns, totaling 23,880 model responses. We design Ben-Util, a checklist-based metric that evaluates how well each model response fulfills the user's benign information need using atomic items. At turn one, models fulfill only 10.5--37.6% of the user's benign information need. When the same query includes the benign intent upfront, models fulfill 25.1--72.1%, confirming that models withhold information due to intent misinterpretation, not limited knowledge. With benign clarifications in multi-turn conversations, 13 of 14 models approach or exceed this single-turn baseline, yet recovery cost varies across models. We identify three failure modes invisible to single-turn evaluations: utility lock-in, where a model rarely updates despite clarification; unsafe recovery, where a model updates at disproportionate safety cost; and repetitive recovery, where a model recycles prior responses rather than providing new information. Moreover, conversations converge to similar harmfulness levels regardless of how conservative the model starts. These findings expose a gap that single-turn evaluations miss -- whether a model is appropriately cautious or simply unresponsive to clarified user intent.

Related papers

• One More Turn, Less Regret: A Regret-Based Multi-Turn Benchmark for LLMs' Clarification Policies [49.067670811577045]
  We introduce RegretBench, a benchmark that evaluates clarification as policy behavior rather than isolated question quality.<n>RegretBench measures intent resolution, interaction cost, ineffective clarification, and regret.<n>Our results show that effective clarification requires more than plausible questions.
  arXiv  Detail & Related papers  (2026-07-23T10:22:11Z)
• Can LLMs Reliably Self-Report Adversarial Prefills, and How? [9.80193616788089]
  We show that large language models (LLMs) exhibit introspective capability on benign tasks.<n>We examine how reliably a model can recognize that its own prior response was elicited by an adversarial prefill attack.
  arXiv  Detail & Related papers  (2026-06-22T17:56:30Z)
• Efficient Safety Benchmarking via Item Response Theory [0.4036506515685838]
  We analyze a suite of widely used safety benchmarks and make three contributions toward more efficient safety evaluation.<n>We show that Item Response Theory (IRT) recovers interpretable structure on safety benchmarks, with ability estimates resolving differences among models.<n>Second, we show that adaptive item selection, which dynamically chooses informative items for each model based on its responses, approximates full-benchmark rankings.<n>Third, we introduce a practical procedure for extracting a fixed, informative subset of items reusable across models, providing an alternative to adaptive selection.
  arXiv  Detail & Related papers  (2026-05-26T17:35:31Z)
• ProactBench: Beyond What The User Asked For [5.422521416406412]
  ProactBench decomposes it into three phase-tied types: textscEmergent, inference from a single disclosed anchor; textscCritical, synthesis across multiple anchors; and textscRecovery, grounded forward-looking value after task completion.<n>Our information asymmetries defend against style-confounding scoring, leakage, external-context contamination, and information dumps.
  arXiv  Detail & Related papers  (2026-05-09T23:56:04Z)
• HuggingR$^{4}$: A Progressive Reasoning Framework for Discovering Optimal Model Companions [50.61510609116118]
  HuggingR$4$ is a novel framework that combines Reasoning, Retrieval, Refinement, and Reflection to efficiently select models.<n>It attains a workability rate of 92.03% and a reasonability rate of 82.46%, surpassing existing method by 26.51% and 33.25% respectively.
  arXiv  Detail & Related papers  (2025-11-24T03:13:45Z)
• Evaluating & Reducing Deceptive Dialogue From Language Models with Multi-turn RL [64.3268313484078]
  Large Language Models (LLMs) interact with millions of people worldwide in applications such as customer support, education and healthcare.<n>Their ability to produce deceptive outputs, whether intentionally or inadvertently, poses significant safety concerns.<n>We investigate the extent to which LLMs engage in deception within dialogue, and propose the belief misalignment metric to quantify deception.
  arXiv  Detail & Related papers  (2025-10-16T05:29:36Z)
• Self-Consistency as a Free Lunch: Reducing Hallucinations in Vision-Language Models via Self-Reflection [71.8243083897721]
  Vision-language models often hallucinate details, generating non-existent objects or inaccurate attributes that compromise output reliability.<n>We present a novel framework that leverages the model's self-consistency between long responses and short answers to generate preference pairs for training.
  arXiv  Detail & Related papers  (2025-09-27T10:37:11Z)
• Reasoned Safety Alignment: Ensuring Jailbreak Defense via Answer-Then-Check [32.82170313959032]
  We introduce a novel safety alignment approach called Answer-Then-Check.<n>Our method enables models to directly answer the question in their thought and then critically evaluate its safety.<n>We find that training on a small subset of just 500 examples can achieve comparable performance to using the full dataset.
  arXiv  Detail & Related papers  (2025-09-15T06:47:35Z)
• FalseReject: A Resource for Improving Contextual Safety and Mitigating Over-Refusals in LLMs via Structured Reasoning [12.467239356591238]
  FalseReject is a comprehensive resource containing 16k seemingly toxic queries accompanied by structured responses across 44 safety-related categories.<n>We propose a graph-informed adversarial multi-agent interaction framework to generate diverse and complex prompts.<n>We show that supervised finetuning with FalseReject substantially reduces unnecessary refusals without compromising overall safety or general language capabilities.
  arXiv  Detail & Related papers  (2025-05-12T20:45:25Z)
• Rethinking LLM Uncertainty: A Multi-Agent Approach to Estimating Black-Box Model Uncertainty [47.95943057892318]
  Quantifying uncertainty in black-box LLMs is vital for reliable responses and scalable oversight.<n>We introduce DiverseAgentEntropy, a novel, theoretically-grounded method employing multi-agent interaction for uncertainty estimation.
  arXiv  Detail & Related papers  (2024-12-12T18:52:40Z)
• SORRY-Bench: Systematically Evaluating Large Language Model Safety Refusal [64.9938658716425]
  SORRY-Bench is a proposed benchmark for evaluating large language models' (LLMs) ability to recognize and reject unsafe user requests.<n>First, existing methods often use coarse-grained taxonomy of unsafe topics, and are over-representing some fine-grained topics.<n>Second, linguistic characteristics and formatting of prompts are often overlooked, like different languages, dialects, and more -- which are only implicitly considered in many evaluations.
  arXiv  Detail & Related papers  (2024-06-20T17:56:07Z)
• MOVE: Effective and Harmless Ownership Verification via Embedded External Features [104.97541464349581]
  We propose an effective and harmless model ownership verification (MOVE) to defend against different types of model stealing simultaneously.<n>We conduct the ownership verification by verifying whether a suspicious model contains the knowledge of defender-specified external features.<n>We then train a meta-classifier to determine whether a model is stolen from the victim.
  arXiv  Detail & Related papers  (2022-08-04T02:22:29Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.