2026-09-09

Refuse without Refusal: A Structural Analysis of Safety-Tuning Responses for Reducing False Refusals in Language Models

👍 4

09/04 08:00

Striking a balance between helpfulness and safety remains a fundamental challenge in aligning large language models. To achieve this balance, models should refuse harmful queries (e.g., "How do I shoot someone?") while remaining responsive to benign inputs, even those superficially resembling harmfu

Beneath the Surface of Chains-of-Thought: A Mechanistic Interpretation of Reasoning Operations in LLMs

👍 14

09/04 08:00

Reasoning in large language models unfolds through diverse functional operations, such as problem formulation, goal decomposition, and deduction. Although these operations are explicitly distinguished in text, little is known about how they are geometrically organized in representation spaces. To th

τ^τ-Bench: An Environment for End-To-End, Realistic Agent Construction

👍 7

09/04 08:00

LLM agents are rapidly becoming production software, deployed to handle customer service, adjudicate disputes, and operate internal systems. Notably, the work of building them is increasingly handed to coding agents, yet existing benchmarks say little about whether an AI system can deliver one under

Ask Before You Optimize: Dynamic Pre-Formulation Clarification for Interactive Optimization

👍 18

09/04 08:00

Large language models (LLMs) are increasingly used to formulate optimization models from natural-language problem descriptions, yet realistic operations research (OR) requests are often incomplete: missing objectives, constraints, or business rules can change the resulting mathematical program. Exis

RISE: Recursive Improvement via Self-Extrapolating Policy Distillation

👍 10

09/04 08:00

On-policy distillation (OPD) provides dense, per-token supervision for language model post-training, but its effectiveness is bottlenecked by teacher quality: external teachers suffer from distribution mismatch, while self-distillation with privileged conditioning is limited by in-context learning c

Privacy Failure in Split-LLM Training, The Returned Gradient Nullifies the Decoys

👍 3

09/03 08:00

We present a systems-security case study of a two-node split-LLM training system whose privacy evaluation passed while leaving an observable channel untested. The Trusted Local Node (TLN) sends protected activations to the Untrusted Cloud Node (UCN), the UCN returns its output, and TLN, holding the

Safety for Whom? Boundary-Aware Self-Distillation for Controlled LLM Safety Refusal

👍 4

09/03 08:00

Safety alignment is usually posed as a topic-level question: is this subject harmful? Deployments ask a narrower one. A civics tutor and a public-sector assistant may share a base model yet need different boundaries inside the same topic, refusing targeted political manipulation while still answerin

ENEAS: Embedding-guided Neural Ensemble for Adaptive Segmentation

👍 35

09/03 08:00

We present ENEAS, a unified, text-promptable method for instance tracking and semantic discovery. Text-promptable segmentation models, including the latest foundation models such as SAM 3, still suffer from temporal hallucinations, spatial fragmentation, and semantic misclassification: they fail to

Unlocking Lossless Speedups in LLMs via Discrete Diffusion

👍 106

09/03 08:00

Large Language Models (LLMs) owe much of their success to next-token prediction (NTP), but their autoregressive (AR) structure requires slow, sequential token generation. To overcome this bottleneck, we introduce diffusion-augmented LLMs, a new class of models that defines an AR model distribution w

FlowBalance: Verifier-Grounded Self-Improvement from On-Policy Reasoning Experience

👍 81

09/03 08:00

A reasoning model can improve from its own on-policy experience, but this inner loop is fragile: terminal verifiers provide reliable yet sparse supervision, while dense same-model guidance can reinforce false confidence or overconcentrate learning on a narrow solution mode. We introduce FlowBalance,

What Else Needs Fixing? Exploring Cost-Effective Test-Time Compute for Revision Propagation in Artifacts Generated Through Conversation

👍 3

09/03 08:00

Large Language Models (LLMs) often help users generate artifacts through iterative cycles of generation and revision in conversation. A challenge here is that, when users specify only a local change during revision, LLMs must instead identify the relevant dependencies and propagate the revision to a

Unfold The World: Factorize 4D Properties in Reinforcing Spatial Reasoning

👍 8

09/03 08:00

Despite the remarkable prowess of Vision-Language Models (VLMs) in general multimodal tasks, they remain fundamentally ``flat'' when reasoning about the physical world. We argue that this spatial bottleneck stems from a profound dimensional mismatch: while VLMs are trained to interpret 2D projection

HarvestBench: Measuring Whether LLM Agents Will Pay to Avoid Killing Animals

👍 3

09/03 08:00

Benchmarks for the side effects an agent causes on the way to a goal already exist, but HarvestBench is the first to put a price on avoiding the side effect and to name that side effect as a living creature. It is a farm simulation: LLM sub-agents drive a crew of two tractors through a cooperative c

The 2026 PNPL Competition: Word Classification and Efficient Cross-Subject Generalisation in LibriBrain100

👍 2

09/03 08:00

The ambition of the 2025 PNPL competition (Landau et al., 2025) was to launch a multi-year curriculum for non-invasive speech decoding. Designed to progress from foundational tasks toward the linguistic complexity required for a practical brain-computer interface (BCI), it set the stage with speech

The Attention Triangle in Audio-Video Models

👍 29

09/03 08:00

Audio-video diffusion models rely on cross-modal attention to coordinate text, sound, and visual content, yet this same mechanism can introduce subtle and systematic semantic leakage. We study these models by probing and analyzing the ``attention triangle,'' comprising the three cross-attention edge

When Models Edit Too Much: On the Fidelity of Minimal Code Edits

👍 7

09/03 08:00

Large language models (LLMs) are increasingly used to edit existing code, but correctness alone is not enough: useful repairs should also be minimal, reviewable, and faithful to the original implementation. We study over-editing, the tendency of a model to rewrite code beyond what is required to fix

MaxKernel: Agentic Kernel Generation for TPUs

👍 12

09/03 08:00

Designing and authoring high-performance custom kernels for accelerators is a complex task that requires deep hardware-level expertise. Large Language Models (LLM) can be leveraged together with real-time compiler feedback to build agentic systems for kernel generation. In this work, we present MaxK

Unifying Conformal Language Tasks with In-Context Ensembles

👍 5

09/02 08:00

Many NLP tasks, such as summarization and extractive question answering, reduce to retrieving relevant content from documents under two constraints: coverage, retaining enough pertinent information to achieve some goal, and conciseness, removing as much irrelevant information as possible. Conformal

Causal Foundation Models

👍 11

09/02 08:00

Causal inference is the practice of estimating the effect of a treatment or intervention from data. It traditionally requires a bespoke pipeline for every new problem: first proposing a causal mechanism, selecting a compatible estimator, and finally training it. Meanwhile, across diverse settings an

Verify Before You Distill: Prompt-Level Teacher Gating for On-Policy Distillation

👍 10

09/02 08:00

On-policy distillation (OPD) accelerates post-training by providing dense token-level supervision from a frozen teacher on the student's own rollouts. Vanilla OPD applies this supervision uniformly across prompts, without checking whether the teacher is reliable for each prompt. Because reverse KL i

Bilevel Coordinated Reflection: A Game-Theoretic Approach to Multi-Agent LLM Systems

👍 130

09/02 08:00

Multi-agent LLM systems commonly use an orchestrator to decompose a task for a team of workers and then improve through textual reflection. Despite strong empirical results, these systems lack a unified account of coordination, memory improvement, and the role of external verification. We model orch

ShallowStream: Index Shallow then Answer Deep for Streaming Video Understanding

👍 3

09/02 08:00

Streaming video understanding is a critical capability for real-world applications, including embodied intelligence, autonomous driving, industrial monitoring, surveillance and early warning, and wearable assistants. However, processing continuous video streams with multimodal large language models

Enoki: Efficient Multi-Level Hallucination Detection

👍 23

09/01 10:21

Ensuring factuality remains a critical challenge for deploying LLMs in high-stakes settings. Existing hallucination detectors usually operate at a single level: claim-level methods provide interpretable factual units, while span-level methods localize unsupported text. Bridging these views is costly

EmbodiedSkills: A Unified Framework for Orchestrating, Training, and Deploying VLA Agents

👍 10

09/01 08:00

Vision-language-action (VLA) models map visual observations and language instructions directly to robot actions, but long-horizon tasks require more than action prediction. An agent must coordinate perception, planning, execution, progress verification, and recovery as the physical state evolves. An

Dr. Claw: An AI Scientist Workspace for Vibe Research

👍 125

08/31 08:00

Command-line coding agents (e.g., Claude Code, Gemini CLI) can already read and write files and sustain long sessions, yet end-to-end research still fragments across chat tools, IDEs, terminals, and writing environments, and the decisions that make it auditable are rarely preserved. We present Dr. C

To See a World in a Living Context: Unified Indoor-Outdoor Urban World Generation

👍 6

08/29 08:00

Text-driven 3D generation has advanced rapidly in creating large-scale outdoor environments and detailed indoor scenes, but these domains are usually synthesized independently, lacking the correspondence required for a coherent urban world. We present HoloWorld, a unified indoor-outdoor urban world

Motion-Omni: End-to-End Joint Speech and Full-Body Motion for Spoken Dialogue

👍 38

08/28 08:00

An avatar that holds a conversation should decide what to say and to move while saying it, yet these abilities live in separate model families: spoken dialogue models produce speech without motion, and co-speech motion models produce motion only from audio handed to them. The standard remedy is a ca

One Symptom, Three Levers: A Critical Review of On-Policy Self-Distillation

👍 10

08/26 08:00

On-policy distillation trains a language model on its own generations while a teacher scores them token by token. It combines the dense supervision of imitation learning with the on-policy sampling of reinforcement learning. But it requires a second, larger model to act as teacher. On-Policy Self-Di

Real-World Knowledge-Guided Change Data Synthesis for Remote Sensing

👍 3

08/25 08:00

Change data synthesis provides a cost-effective solution for expanding training data and improving the performance of change detection models. However, existing synthesis methods typically rely on handcrafted rules to simulate changes, where limited coverage of class transitions restricts the divers

Training-Free Speech-Centric Omni Understanding with Frozen VLMs

👍 4

08/07 08:00

Audio-visual understanding remains challenging because models must jointly interpret spoken content, visual events, and their temporal relationships. Existing omni models typically introduce dedicated audio encoders and rely on expensive audio-video-text training, tightly coupling omni capability to

GPT-6 Astra should be Illegal for SEO

@borjafat · 4.9K 粉丝 · 373.2K 阅 · 502 赞 · 42 转

09/07 19:52

If you want more customers from Google and AI search, but: ↳ ChatGPT keeps giving you keyword lists while your existing pages still need work. ↳ You open Search Console, stare at the numbers, and

中文介绍 分析GPT-6 Astra对SEO的影响，指出其提供关键词列表而忽略现有页面优化的问题。

Claude Code → Codex (and keep everything you already built)

@charliejhills · 14.1K 粉丝 · 120.8K 阅 · 504 赞 · 53 转

09/06 18:44

For months people have been telling me Codex is better than Claude Code. I ignored all of it. I have built my whole business inside Claude Code (every pipeline, every system) and I did not fancy

中文介绍 介绍如何迁移从Claude Code到Codex，保留现有业务构建。

Reducing cost and improving performance with Claude Platform

@ClaudeDevs · 683.4K 粉丝 · 106.4K 阅 · 1.3K 赞 · 70 转

09/09 01:01

Tuning prompt caching, instructions, and effort can reduce Claude's cost without sacrificing application performance. Performance and cost are often viewed as a trade-off: to spend less, you accept

中文介绍 讨论如何调整Claude平台参数以降低成本并提升性能。

How to Use GPT-6 Astra to Find Profitable Trading Strategies 24/7

@RohOnChain · 68.8K 粉丝 · 102.3K 阅 · 524 赞 · 48 转

09/08 20:39

I will break down exactly how to build an AI agent like hedge funds that finds profitable trading strategies from the markets 24/7 with GPT-6 Astra. Let's get straight to it. Bookmark This - I'm

中文介绍 讲解如何使用GPT-6 Astra构建全天候寻找交易策略的AI代理。

Using Grok Bot: Figma Photomosaics

@mattyp · 60.1K 粉丝 · 82.8K 阅 · 515 赞 · 56 转

09/08 06:03

You may have seen some three-panel mosaics on your timeline. Here's how you can create them with @bot in @figma. Any designer knows how they'd make one, it's just a bit of a pain in the ass. So we

中文介绍 展示如何使用Grok Bot在Figma中创建照片拼贴。

Exactly how I edit videos with Astra

@ForwardEditor · 10.0K 粉丝 · 61.4K 阅 · 512 赞 · 36 转

09/08 08:54

This weekend, I watched a movie with my daughter while my computer worked on a video I would have normally spent most of a day editing. The video was a favor for someone in my personal life. I'm a

中文介绍 分享使用Astra编辑视频的经验，实现快速视频编辑。

The Math Needed for AI Inference Engineering (Complete Roadmap)

@TheVixhal · 23.0K 粉丝 · 59.7K 阅 · 541 赞 · 58 转

09/08 01:07

In this article, I'm going to break down the essential math you need for AI inference engineering. I'll also share the exact roadmap and resources that helped me personally. Let's get straight to it.

中文介绍 概述AI推理工程所需的数学基础及个人学习路径。

Grok Bot Marketplace is live!

@ericzakariasson · 87.8K 粉丝 · 58.2K 阅 · 522 赞 · 49 转

09/07 19:05

We launched the Grok Bot Marketplace last week on Grok Bot · Bot Marketplace! It's a shelf of teammates other people already built for real jobs. Pick one, import it into your sidebar, and you're

中文介绍 介绍Grok Bot市场上线，提供他人已构建的团队助手。

AlphaGenome Atlas: A High-Resolution Map of Human DNA Accelerating Biological Discovery

@GoogleAI · 2.5M 粉丝 · 37.4K 阅 · 643 赞 · 90 转

09/08 22:41

Today, we’re introducing AlphaGenome Atlas 🧬 an interactive platform for genomics discovery. In 2025, @GoogleDeepMind released AlphaGenome to help decode the complex language of our DNA — but

中文介绍 发布AlphaGenome Atlas，一个用于基因组发现的交互式平台。

How GPT-5.6 Sol helps run quantum computing experiments

OpenAI News · 09/09 01:00

See how an MIT researcher uses GPT-5.6 Sol with Codex to autonomously run quantum computing experiments, analyze results, and calibrate qubits.

中文介绍 MIT研究人员利用GPT-5.6 Sol与Codex自主运行量子计算实验，分析结果并校准量子比特。

Safety for Whom? Refusing the Right Subset of a Topic, Not the Whole Topic

Hugging Face Blog · 09/08 22:23

中文介绍 探讨拒绝话题的子集而非整个话题的安全性问题。

AlphaGenome Atlas: A predictive map of every possible DNA letter change in the human genome

DeepMind Blog · 09/08 22:00

AlphaGenome Atlas maps the molecular effects of 9 billion single-letter DNA variants across the human genome.

中文介绍 AlphaGenome Atlas绘制了人类基因组中所有可能的DNA字母变化的分子效应图。

The Work Now Within Reach

OpenAI News · 09/08 21:00

Explore how more capable, affordable AI can expand the work people and businesses can accomplish—and make growth more economical.

中文介绍 探索更强大、更具成本效益的AI如何扩展人们和企业能完成的工作，并使增长更经济。

Introducing ChatGPT Images 2.5

OpenAI News · 09/08 19:30

ChatGPT Images 2.5 helps turn your ideas, sketches, and reference photos into more personalized, polished images that better reflect your ideas.

中文介绍 ChatGPT Images 2.5帮助将您的想法、草图和参考照片转化为更个性化、更完美的图像。

This AI entrepreneur is developing agents that can plan ahead for the unexpected

MIT Tech Review AI · 09/08 18:34

Danijar Hafner’s office in San Francisco’s SoMa district sits mostly empty. His brand-new startup is still in stealth mode and doesn’t even have its name on the door. On the day I visit, there’s only one other person there, and little in the way of furniture. But what it lacks in decor, it makes up…

中文介绍 AI企业家丹尼贾尔·哈夫纳正在开发能够为意外情况做出计划的智能代理。

On the Navier–Stokes Millennium Prize Problem

OpenAI News · 09/08 18:00

We’re sharing an AI-generated solution to the Navier–Stokes Millennium Prize Problem, including a writeup and a formal proof in Lean.

中文介绍 分享了对Navier-Stokes千年大奖问题的AI生成解决方案，包括Lean中的说明和正式证明。

Funding grants for new research into AI and teen development

OpenAI News · 09/08 17:00

Apply now for OpenAI’s $5 million grant program supporting independent research on how generative AI affects teen development, well-being, and safety.

中文介绍 OpenAI推出500万美元的补助金计划，支持独立研究生成式AI对青少年发展、福祉和安全的影响。

OpenAI expands initiatives to support journalism from classrooms to newsrooms

OpenAI News · 09/08 08:00

OpenAI is expanding support for journalism with tools, training, and partnerships for students, educators, journalists, and news organizations.

中文介绍 OpenAI通过工具、培训和合作伙伴关系扩展对新闻业的支持，从课堂到新闻室。

1Password increases engineering productivity 21% with Codex

OpenAI News · 09/08 08:00

Engineers at 1Password use Codex to rapidly build new features and internal tools, reaching production-readiness while maintaining rigorous security policies.

中文介绍 1Password工程师使用Codex快速构建新功能和内部工具，同时保持严格的安全政策。

OpenAI managed agents 🤖, TPU inference ⚡, Anthropic $517B compute 💰

TLDR AI · 09/08 08:00

中文介绍 OpenAI管理代理、TPU推理、Anthropic 517亿美元计算。

The Frontier AEO Tracker: What Astra Chooses (and every other frontier model, and what you can do about it)

Latent Space · 09/08 05:32

Our first Astra project dives into AEO trends, a top asked topic from founders and DX leaders we talk to.

中文介绍 AEO前沿追踪器：Astra的选择以及每个前沿模型的选择，以及您可以做什么。

Supporting independent journalism in Ukraine

OpenAI News · 09/07 08:00

OpenAI, AIRPPU and WAN-IFRA launch an AI program to help Ukrainian news organizations strengthen innovation, resilience, and independent journalism.

中文介绍 OpenAI、AIRPPU和WAN-IFRA启动一个AI项目，帮助乌克兰新闻机构加强创新、韧性和独立新闻。

Claude proves Fermat 🧮, automated AI researcher 🔬, Z1 efficiency chip ⚡

TLDR AI · 09/07 08:00

中文介绍 Claude证明费马定理、自动化AI研究员、Z1效率芯片。

An Alien Mind

OpenAI News · 09/06 17:00

Jakub Pachocki reflects on increasingly capable AI and the challenge of keeping it aligned. He calls for stronger safeguards and international coordination.

中文介绍 雅库布·帕乔茨基反思越来越强大的AI以及保持其一致性的挑战。他呼吁更强大的保障和国际协调。