精选

COPE：利用用户嵌入与自评估实现大模型在稀疏反馈下的持续个性化

针对大语言模型在标准对齐后输出同质化、难以满足用户多样化偏好的问题，研究人员提出了COPE优化框架。现有免训练方法多依赖提示工程占用上下文窗口，而微调方法在训练后保持静态，无法持续适应动态偏好。COPE结合个性化用户嵌入与模型自评估机制，有效克服了现实场景中用户反馈稀疏的挑战，实现了大语言模型在部署后的持续个性化优化。

QUARTET：基于四分支交叉注意力与随机游走轨迹增强的关系图Transformer

关系深度学习将多表数据库建模为异构时序图，目前图Transformer在该领域表现优异。然而，现有代表性模型RelGT存在局部采样子图连接松散阻碍消息传递，以及全局注意力依赖单一特征记忆而忽视宏观动态的两大缺陷。为此，研究人员提出了QUARTET架构，通过结合四分支交叉注意力与随机游走轨迹，增强全自注意力机制在关系图表征中的表达能力。

面向网络生成文本的字素转音素基准 UGTPhon：涵盖多语言与组合式方法

文本转语音（TTS）系统在处理网络生成文本（如缩写 ppl、imo）时，需根据规范表达而非表面拼写来推断发音。为此，研究人员推出了首个针对英语、越南语和韩语网络文本的字素转音素（G2P）基准 UGTPhon，并配套提出用于细粒度诊断的分类体系。测试显示，现有 G2P 模型及前沿大模型在此类文本上的音素错误率（PER）最高落后 66.8 点。研究团队进一步提出了一种结合规范形式证据的组合式 G2P 基线方法。

智能体训练中工具缓存可能导致策略更新方向反转

工具结果缓存可显著减少智能体训练中的重复计算，但也会耦合采样过程的随机性。最新研究通过一个双动作模型证明，即使独立执行与共享执行在边际上保持了每个采样的条件奖励分布一致，在组内共享随机结果仍可能导致预期的组归一化策略更新方向发生反转。理论推导表明，共享缓存更新反映的是胜负概率之差而非期望奖励之差，揭示了智能体强化学习中常见工程优化策略潜在的算法偏差风险。

LOCKR：基于隐状态轨迹引导检测与修复扩散语言模型的推理锁定问题

扩散语言模型通过迭代去噪生成文本。研究团队发现该类模型中存在一种“稳定但错误锁定”的推理故障，即答案在去噪阶段早期便过早稳定于错误结果。置信度、熵等常规表层解码信号难以有效区分正误锁定。为此，论文将选择性推理修复建模为轻量级测试时规划任务，并提出了 LOCKR，一种基于隐状态轨迹引导的规划器，专门用于检测与修复该问题。

Gemini 3.8 text-to-speech says hello

Introducing MentalHealthBench

MentalHealthBench is an expert-informed benchmark for evaluating helpful and safe AI responses across realistic mental health conversations.

Claude 5.5 发布，性能直逼 Fable，还要卷价格

最卷一夜！ #欢迎关注爱范儿官方微信公众号：爱范儿（微信号：ifanr），更多精彩内容第一时间为您奉上。

⁡⁤‬‍⁤‬​‌‍⁢⁡⁢⁣‌‌⁡⁡​⁤​​⁢⁡‬⁡‬​‍⁣​​⁤⁡‍​⁡‬‌‬⁣⁡⁣​⁡‬⁤刚刚，GPT-6 新模型掀桌！「白菜价」杀进 DeepSeek 腹地

DeepSeek 的价格护城河，迎来 OpenAI 的挑战 #欢迎关注爱范儿官方微信公众号：爱范儿（微信号：ifanr），更多精彩内容第一时间为您奉上。

Claude discovers a novel enzyme system with CRISPR-like repeats

Bring more intelligence to everyday work with GPT-6 Sol and GPT-6 Luna on Amazon Bedrock

GPT-6 Sol and GPT-6 Luna are now generally available on Amazon Bedrock, giving you more options to match intelligence and efficiency to each workload.

Introducing GPT-6 Sol and Luna

Meet GPT-6 Sol and Luna, two models that bring frontier intelligence to everyday work with different balances of capability and cost.

Claude Opus 5.5 is now available on AWS

Claude Opus 5.5, Anthropic's most capable Opus model for agentic coding, knowledge work, and long-running tasks, is now available on Amazon Bedrock and Claude Platform on AWS. This post covers what's new in Opus 5.5, practical guidance, and how to start building with the model on Amazon Bedrock.

v0.30.0

v0.30.0 Highlights This release features 762 commits from 315 contributors (104 new)! New models : DeepSeek-V4.1-Flash ( #56214 , #56228 , #56208 ) with the whole KV stored in MXFP8 through the FlashMLA V4.1 record on SM100 ( #56893 ), DeepGEMM Mega-mHC ( #56962 ), and async Engram prefetch with Engram DP sharding ( #56512 ); DeepSeek-V4-Flash-Vision-Exp ( #54566 ), also on ROCm ( #55107 ) and with LoRA ( #55897 ); GLM-5.3-Flash ( #53906 ) with EPLB ( #55119 ); K2-Horizon ( #55063 ); Cohere Compass ( #54774 ); Bailing V3 VL ( #55921 ); Nanbeige4.2 via the Transformers backend ( #56071 ); and a DeepSeek-V4 CPU backend with AVX512/AMX sparse ML

How UK AISI and EvalEval Are Making Benchmark Results Reproducible

Transformers now runs llama.cpp quants

Introducing Claude Opus 5.5

xAI’s Grok 4.6 is now available in Amazon Bedrock

xAI's Grok 4.6 is now available in Amazon Bedrock: a frontier model for long-running agents, coding, and knowledge work, with a 500K token context window and four reasoning effort levels. It runs on both the bedrock-mantle and bedrock-runtime endpoints, with Converse API and cross-Region inference support.

Improving synthesis prediction of small molecules at scale with RetroChimera

Custom-made molecules are advancing medicine, materials, and agriculture, but producing them is slow and expensive. A new Nature paper highlights RetroChimera, a predictive model that helps accelerate chemical synthesis, helping researchers explore a wide range of molecules. The post Improving synthesis prediction of small molecules at scale with RetroChimera appeared first on Microsoft Research .

Pruning LLMs Like a Physicist: Block Removal as an Ising Optimization Problem

tokenizers v1: encode, decode and scaling, measured

MilleMiglia: A realistic instance generator for middle-mile logistics

Algorithms & Theory

Partnering with Accenture on embedded evaluation

Introducing the Life Sciences Verification Program

University of Manchester Uses NVIDIA Earth-2 to Forecast Air Pollution Across the UK

Air pollution is a serious public health risk, contributing to an estimated 30,000 deaths in the U.K. alone last year. Data-driven insights can help — but computing air quality with traditional chemistry-based models is expensive, which limits how detailed they can be and how regularly they can be run. David Topping, a professor in the […]

Bypassing inference bottlenecks: Accelerating complex AI search with Retrieve-for-Train

Algorithms & Theory

Introducing Gemini 3.8 Live and 3.8 Live Extended Thinking

AI for everyone in every language

Animation of several words in different languages slowly zooming past

Heart of the Matter: How a Major Children’s Hospital Uses Open Source NVIDIA AI for Cardiac Care

ToolGrad: Efficient tool-use dataset generation with textual "gradients"

Machine Intelligence