1
腾讯混元 Hy Image3.5 preview 发布：支持文生图与多轮对话创作，已上线元宝、ima 等应用
IT之家 9 月 22 日消息，腾讯混元团队今日宣布推出 Hy Image3.5 preview 模型。腾讯称该模型面向专业级生图创作，支持文生图、图生图和多轮对话创作。 腾讯称，该模型经过内部数百名专业设计师 GSB 盲测，对比 Hy Image3.0 提升 30%，与 Seedream 5.0 …
2
Do LLMs Know What to Ask and When? Evaluating Multi-Turn Information Seeking
大模型能否主动追问缺失信息？这项研究评估了多轮对话中的信息寻求能力，为AI“懂不懂装懂”给出答案。
arXiv:2608.14808v1 Announce Type: new Abstract: When a user question is underspecified, a capable model should recognize that its context is insuffici…
3
比亚迪腾势 Z9S 官宣晨曦紫外观 × 云烟紫内饰，还将带来“多巴胺车色”
用AI快速抓取文章核心，支持多轮对话追问细节，还能一键生成摘要和对比表格
IT之家 7 月 28 日消息，继此前公布“晨曦紫”外观后，今天上午，比亚迪旗下腾势汽车正式揭晓了与之相呼应的腾势 Z9S“云烟紫”内饰配色，并在官微中提到新车将拥有一系列“多巴胺车色”。 外观细节上，新车配有 车顶激光雷达、智驾“小蓝灯”、半隐藏式门把手以及可升降尾翼 ，车尾有望配备可点亮的车标。…
4
LLMs Get Lost in Evolving User Intent
研究发现大型语言模型在用户意图动态演变时容易“迷失”，揭示了多轮对话中推理与记忆的脆弱性。
arXiv:2607.20734v1 Announce Type: new Abstract: As LLMs become more capable, they are increasingly deployed as collaborative agents, taking on user-de…
5
Stateful Guardrails for Multi-Turn LLM Systems: A Conversational Risk Accumulation Framework
多轮LLM对话安全的新框架，通过状态跟踪实现风险累积防护，应对长对话中的隐性威胁。
arXiv:2607.19361v1 Announce Type: new Abstract: Most safety guardrails for large language models (LLMs) evaluate each prompt-response pair in isolatio…
6
RSPO: Reward-Swap Policy Optimization for Multi-Turn LLM Agents
提出Reward-Swap策略优化，解决多轮对话LLM代理奖励稀疏问题，提升推理与交互效率。
arXiv:2607.04713v1 Announce Type: cross Abstract: Reinforcement learning holds significant potential for training large language models (LLMs) to hand…
7
How Many Iterations to Jailbreak? Dynamic Budget Allocation for Multi-Turn LLM Evaluation
提出动态预算分配方法，高效评估多轮对话中LLM的越狱风险，解决计算瓶颈与稀有事件检测难题。
arXiv:2605.06605v2 Announce Type: replace Abstract: Evaluating and predicting the performance of large language models (LLMs) in multi-turn conversati…
8
DiPS: Dialogue Policy Selection for High-Stakes Persuasion Agents
面向高风险场景的说服智能体，提出对话策略选择新方法，值得关注。
arXiv:2607.01557v1 Announce Type: cross Abstract: Large Language Models (LLMs) often struggle with persuasion in high-stakes scenarios. People's indiv…
9
BayesBench: Evaluating LLM Belief Trajectories Under Multi-Turn Evidence Accumulation
多轮对话中模型信念如何随证据更新？新基准BayesBench专测LLM贝叶斯推理轨迹
arXiv:2606.30850v1 Announce Type: new Abstract: Large language models (LLMs) are typically deployed in multi-turn conversations, where each turn provi…
10
Accuracy and Satisfaction in Multi-Turn LLM Dialogues for NFR Assessment
首次聚焦多轮LLM对话中非功能性需求评估，揭示准确性与用户满意度的新挑战
arXiv:2606.24834v1 Announce Type: new Abstract: LLM-based dialogue assistants have become mainstream tools for software developers, yet current evalua…
11
LLM agent safety, multi-turn red-teaming, jailbreak benchmarks, adversarial robustness, safety-critical systems
首个聚焦安全关键控制室的LLM操作员多轮红队基准，覆盖对抗鲁棒性与越狱攻击评测。
arXiv:2606.20408v1 Announce Type: cross Abstract: Large language model (LLM) agents are increasingly proposed as supervisory components for safety-cri…
12
EHRNote-ChatQA: A Benchmark for Evidence-Grounded Multi-Turn Clinical Question Answering over Longitudinal Discharge Summaries
面对长篇出院小结进行多轮查询时，Perplexity AI 能基于检索增强生成提供带来源的证据回答，助力医学专家高效合成病历信息。
arXiv:2606.15735v1 Announce Type: cross Abstract: Discharge summaries are crucial clinical documents containing the context of a patient's overall hos…
13
P3B3: A Multi-Turn Conversational Benchmark for Measuring European and Brazilian Portuguese Variety Bias in LLMs
专门评估大语言模型对欧式与巴西葡萄牙语变体偏见的首个多轮对话基准P3B3。
arXiv:2606.16753v1 Announce Type: cross Abstract: As Large Language Models (LLMs) become embedded in everyday communication, capturing regional lingui…
14
How to Build an AI Agent for Slack with Chat SDK and AI SDK
用Vercel AI SDK和Chat SDK轻松搭建自带流式响应和工具调用的Slack AI Agent
Article URL: https://vercel.com/kb/guide/how-to-build-an-ai-agent-for-slack-with-chat-sdk-and-ai-sdk Comments URL: https://news.ycombinator.com/item?i…
15
PsychoPass: Geometric Profiling of Multi-Turn Adversarial LLM Conversations
用几何方法剖析多轮对话中LLM的对抗性攻击模式，为AI安全提供全新视角
arXiv:2606.03136v1 Announce Type: cross Abstract: Multi-turn jailbreak attacks on large language models (LLMs) reveal a mismatch in current guardrails…
16
Evolving and Detecting Multi-Turn Deception using Geometric Signatures
利用几何签名方法检测多轮对话中的欺骗行为，为AI安全提供新思路。
arXiv:2605.27671v1 Announce Type: cross Abstract: Safety defenses for large language models (LLMs) are typically trained and evaluated on single-turn …
17
SaFeR-Steer: Evolving Multi-Turn MLLMs via Synthetic Bootstrapping and Feedback Dynamics
提出SaFeR-Steer方法，通过合成引导和反馈动态机制进化多轮多模态大模型，创新模型训练范式。
arXiv:2604.16358v2 Announce Type: replace Abstract: MLLMs are increasingly deployed in multi-turn settings, where attackers can escalate unsafe intent…
18
Stop Listening to Me! How Multi-turn Conversations Can Degrade LLM Reliability
多轮对话让LLM可靠性下降？研究揭示在医疗等高风险场景中静态评测与真实使用的巨大鸿沟。
arXiv:2603.11394v3 Announce Type: replace-cross Abstract: Large language models (LLMs) excel on static benchmarks, but their performance across multi-…
19
AIDG: A Formal Decomposition of Information Extraction and Containment Asymmetries in Multi-Turn LLM Dialogue
多轮对话中信息不对称的数学建模，为LLM安全交互提供新理论框架
arXiv:2602.17443v2 Announce Type: replace Abstract: Multi-turn LLM evaluation is typically reported as a single win-rate scalar, conflating distinct c…
20
Memory Architectures for Multi-Turn Text-to-SQL: A Benchmark and Empirical Study
多轮对话中记忆架构如何影响Text-to-SQL性能？基准与实证研究揭示关键设计差异。
arXiv:2605.26394v1 Announce Type: new Abstract: Multi-turn Text-to-SQL is central to enterprise analytics yet remains predominantly evaluated in singl…