model releases

The Geometry of Representational Failures in Vision Language Models

arXiv:2602.07025v2 Announce Type: replace-cross Abstract: Vision-Language Models (VLMs) exhibit puzzling failures in multi-object visual tasks, such as hallucinating non-existent elements or failing t

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Model Releases

The highest leverage work in AI right now is some of the most boring. (Well boring to others, I kind of love the pain of problem-solving.) E…

DGX agent

The highest leverage work in AI right now is some of the most boring. (Well boring to others, I kind of love the pain of problem-solving.) Everyone and their boss wants to build the cool AI agent... t

model-releasesallie-k--miller--x

8 Jun 2026

The north stars we're working towards at OpenAI all center around the mission: ensure AGI benefits all of humanity. AI should expand human a…

The north stars we're working towards at OpenAI all center around the mission: ensure AGI benefits all of humanity. AI should expand human agency, not make people less consequential to the future. htt

model-releasesopenai--x

The Piggyback Hypothesis of Generalization: Explaining and Mitigating Emergent Misalignment

arXiv:2606.06667v1 Announce Type: new Abstract: The mechanisms behind LLMs' broad over-generalization beyond training examples remain unclear. Emergent misalignment (EM) offers a striking case study:

The Post-GCN Decade Revisited: Curvature-Stratified Evaluation of Relational Learning

arXiv:2606.06397v2 Announce Type: replace Abstract: Current evaluation practices in relational learning rely heavily on flat leaderboards that average performance across heterogeneous datasets, implic

There’s so much demand for a good small model, look at top downloaded qwen models All < 9b

model-releasesclem-delangue--x

The post highlights strong market demand for efficient small language models under 9 billion parameters, citing Qwen's top-downloaded models as evidence of user preference for compact, resource-effici

Think Fast: Estimating No-CoT Task-Completion Time Horizons of Frontier AI Models

arXiv:2606.07157v1 Announce Type: new Abstract: Many efforts to ensure frontier AI models are safe rely on monitoring their chain-of-thought (CoT) reasoning. If models become able to perform sufficien

Think Like a Pilot: Fine-Grained Long-Horizon UAV Navigation

arXiv:2606.06836v1 Announce Type: cross Abstract: Language-guided UAV agents must execute long-horizon semantic instructions while producing smooth, physically feasible continuous flight commands, yet

ThinkBooster: A Unified Framework for Seamless Test-Time Scaling of LLM Reasoning

arXiv:2606.06915v1 Announce Type: cross Abstract: Test-time compute (TTC) scaling has emerged as a powerful paradigm for improving large language model (LLM) reasoning by allocating additional compute

This is super big I think this is the first useful speculative decoding method deployed on a big quasi frontier model Massive unlock @fi5662…

model-releasesjeremy-howard--x

This is super big I think this is the first useful speculative decoding method deployed on a big quasi frontier model Massive unlock @fi56622380 🚀 1,000+ TOKENS/S ON A 1T MODEL! 🚀 We are thrilled to r

TokaMind: A Multi-Modal Transformer Foundation Model for Tokamak Plasma Dynamics

arXiv:2602.15084v2 Announce Type: replace-cross Abstract: We present TokaMind, to our knowledge the first open-source foundation model for tokamak plasma dynamics, based on a Multi-Modal Transformer (

Tokyo this week

model-releasesboris-cherny--x

This post likely provides weekly updates, events, or recommendations for activities and happenings in Tokyo for the current week. The content probably covers entertainment, dining, cultural events, or

Towards Tight Bounds for Streaming Attention

arXiv:2606.07205v1 Announce Type: cross Abstract: The attention mechanism is a cornerstone of modern transformer architectures. However, its expressive power comes at the cost of quadratic runtime and

Trading Engagement for Sustainability: Carbon-Aware Re-ranking for E-commerce Recommendations

arXiv:2606.04550v1 Announce Type: cross Abstract: E-commerce recommender systems strongly influence which products users consider and purchase, yet sustainability signals such as Product Carbon Footpr

Tree-of-Experience: A Structured Experience-Management Solution for Self-Evolving Agents under Low-Repetition and Implicit-Reward Environments

arXiv:2606.06960v1 Announce Type: new Abstract: Experience-based self-evolution is crucial for LLM agents, but existing benchmarks often assume explicit goals, stable task patterns, and clear feedback

TSAQA: Time Series Analysis Question And Answering Benchmark

arXiv:2601.23204v2 Announce Type: replace Abstract: Time series data are integral to critical applications across domains such as finance, healthcare, transportation, and environmental science. While

Twin: Tuning Learning Rate and Weight Decay of Deep Homogeneous Classifiers without Validation

arXiv:2403.05532v2 Announce Type: replace-cross Abstract: We introduce Tune without Validation (Twin), a simple and effective pipeline for tuning learning rate and weight decay of homogeneous classifi

Uncertainty-Aware LLM-Guided Policy Shaping for Sparse-Reward Reinforcement Learning

arXiv:2606.06673v1 Announce Type: new Abstract: Sparse rewards and heterogeneous task sequences remain persistent challenges in Reinforcement Learning (RL), often resulting in slow convergence, weak g

Uncertainty-Guided Label Rebalancing for CPS Safety Monitoring

arXiv:2603.25670v3 Announce Type: replace Abstract: Safety monitoring is essential for Cyber-Physical Systems (CPSs). However, unsafe events are rare in real-world CPS operations, creating an extreme

Uniform Stability and Generalization Error of GD and SGD on Fixed-Point Parameters

arXiv:2606.06934v1 Announce Type: new Abstract: We analyze generalization error, uniform stability, and uniform argument stability of gradient descent (GD) and stochastic gradient descent (SGD) over d

UniSHARP: Universal Sharp Monocular View Synthesis

arXiv:2606.07514v1 Announce Type: new Abstract: In this work, we focus on extending SHARP, the popular photorealistic view synthesis method, for universal monocular rendering across a continuum of cam

UnpredictaBench: A Benchmark for Evaluating Distributional Randomness in LLMs

arXiv:2606.06622v1 Announce Type: new Abstract: We introduce UnpredictaBench, an evaluation that tests the ability of large language models (LLMs) to capture true underlying distributions. As LLMs are

Unsupervised Continual Clustering via Forward-Backward Knowledge Distillation

arXiv:2606.07474v1 Announce Type: new Abstract: Unsupervised Continual Learning (UCL) aims to enable neural networks to learn sequential tasks without labels or access to past data. A major challenge

UrduMMLU: A Massive Multitask Benchmark for Urdu Language Understanding

arXiv:2606.07167v1 Announce Type: cross Abstract: Meaningful multilingual evaluation must test models in the target language and educational context. Urdu, spoken by more than 230 million people, lack

VeriDrive: Verifiable Counterfactual Supervision for Cost-Efficient Vision-Language Planning

arXiv:2606.07338v1 Announce Type: new Abstract: Vision-language driving models increasingly use reasoning supervision to bridge perception, prediction, and planning, but existing driving rationales ar

VideoSEG-O3: A Multi-turn Reinforcement Learning Framework for Reasoning Video Object Segmentation

arXiv:2606.06819v1 Announce Type: new Abstract: Reasoning Video Object Segmentation (RVOS) demands a sophisticated integration of temporal dynamics, spatial details, and linguistic reasoning to achiev

we just shipped support for rubrics in deepagents ✅ give your agent a clear definition of what 'done' looks like, and force it to run in a l…

model-releasesharrison-chase--x

we just shipped support for rubrics in deepagents ✅ give your agent a clear definition of what 'done' looks like, and force it to run in a loop until said goal is complete this is similar to /goal in

What Your Posts Reveal: A Benchmark and Agentic Framework for User-Level Privacy Leakage on Social Media

arXiv:2606.06784v1 Announce Type: cross Abstract: Public social media posts can reveal private information through weak cues scattered across text, images, or metadata. Such leakage is often cumulativ

When Large Language Models Fail in Healthcare: Evaluating Sensitivity to Prompt Variations

arXiv:2606.07237v1 Announce Type: cross Abstract: Large Language Models (LLMs) are increasingly used in healthcare for tasks such as clinical question answering, diagnosis support, and report summariz

When Recovery Matters: The Blind Spot of Surrogate Privacy in MLLM Editing

arXiv:2606.07171v1 Announce Type: new Abstract: Multimodal Large Language Models (MLLMs) enable flexible instruction-driven image editing, but privacy risks arise when user images expose diverse and u

When to Think Deeply: Inhibitory Deliberation for LLM Reasoning

arXiv:2606.06745v1 Announce Type: new Abstract: Reasoning Large Language Models can improve problem-solving performance through deliberative inference, but invoking slow reasoning for every input is c

When we first demoed Claude Code internally, it got two reactions on Slack. A year after GA, @_catwu and I sat down to talk about what's cha…

model-releasesboris-cherny--x

When we first demoed Claude Code internally, it got two reactions on Slack. A year after GA, @_catwu and I sat down to talk about what's changed: why I use auto mode instead of plan mode, how routines

Which Anatomy Matters Under Limited Labels? A Data-Efficient Anatomy-Aware Benchmark for Cardiac Pathology Prediction

arXiv:2606.06509v1 Announce Type: cross Abstract: Numerous medical imaging problems must be solved under limited labels and constrained compute, yet it remains unclear whether performance gains are dr

Wordle 1,814 4/6 ⬛🟨⬛⬛⬛ ⬛⬛⬛⬛⬛ ⬛🟨🟨⬛⬛ 🟩🟩🟩🟩🟩

model-releasesanthropic--x

This post documents a Wordle game result where the player solved puzzle #1,814 in four attempts, using color-coded emoji tiles (⬛ for incorrect letters, 🟨 for correct letters in wrong positions, 🟩 for

WorldBench: A Challenging and Visually Diverse Multimodal Reasoning Benchmark

arXiv:2606.06538v1 Announce Type: new Abstract: In real-world applications, models are expected to perform reliably across diverse settings. Yet, many existing multimodal benchmarks expand task types

Xiaomi claims MiMo-V2.5-Pro-UltraSpeed tops 1K tokens/second, a first at the 1T-parameter scale, using a standard 8-GPU commodity node; API trial starts June 9 (Jose Antonio Lanz/Decrypt)

Jose Antonio Lanz / Decrypt: Xiaomi claims MiMo-V2.5-Pro-UltraSpeed tops 1K tokens/second, a first at the 1T-parameter scale, using a standard 8-GPU commodity node; API trial starts June 9 — Most peop

model-releasestechmeme

Zero-Shot Embedding Drift Detection: A Lightweight Defense Against Prompt Injections in LLMs

arXiv:2601.12359v1 Announce Type: cross Abstract: Prompt injection attacks have become an increasing vulnerability for LLM applications, where adversarial prompts exploit indirect input channels such

A point on the board for 2026. Thank you to the team, our partners, and our fans for the ongoing work and continued support. #MonacoGP

Aston Martin F1 team secured a point at the 2026 Monaco Grand Prix, acknowledging contributions from their team members, partners, and supporters. The post expresses gratitude for the collaborative ef

model-releasescohere--x

datasette-agent-edit 0.1a0

model-releasessimon-willison

Release: datasette-agent-edit 0.1a0 I'm planning several plugins for Datasette Agent which can make edits to existing pieces of text - things like collaborative Markdown editing, updating large SQL qu

Falcon 9 launches 21 @Starlink satellites and two Starshield satellites from California

model-releaseselon-musk--x

SpaceX's Falcon 9 rocket successfully launched 21 Starlink internet satellites and two Starshield military satellites from a California launch facility. Starlink satellites are part of SpaceX's global

Grep timeout issue fixed in latest Grok Build

model-releaseselon-musk--x

Grep timeout issue fixed in latest Grok Build Grok Build update just released v0.2.31 Release Notes: Bug Fixes: • Marketplace skills without proper descriptions are now hidden from listings instead of

Have been extensively testing Claude Workflows this weekend, with the best model possible. Threw it at my whole code base, combing for bugs.…

model-releasesboris-cherny--x

Have been extensively testing Claude Workflows this weekend, with the best model possible. Threw it at my whole code base, combing for bugs. 144 found and fixed! Geez... It is a large code base, for s

LOL. Even Claude sees through Hinton’s nonsense. (though see the articles I posted earlier, for converging sources I put more weight on)

model-releasesgary-marcus--x

LOL. Even Claude sees through Hinton’s nonsense. (though see the articles I posted earlier, for converging sources I put more weight on) This is my prompt and this is Claude's response: ME: What do yo

see eg https://www.scmp.com/tech/tech-trends/article/3271858/ai-race-alibaba-tencent-quickly-adopt-metas-new-llama-31-model-amid-excitement …

model-releasesgary-marcus--x

see eg https://www.scmp.com/tech/tech-trends/article/3271858/ai-race-alibaba-tencent-quickly-adopt-metas-new-llama-31-model-amid-excitement and https://medium.com/the-endless-forge/zuckerbergs-llama-f

The first wave of AI-native applications is wrapping tokens and providing in-app agents. As agent usage centralizes around core apps (e.g. C…

model-releasesjerry-liu--x

The first wave of AI-native applications is wrapping tokens and providing in-app agents. As agent usage centralizes around core apps (e.g. Claude Code, Codex), there's this emerging wave of building s

There was an inflection point recently where the tide shifted to model pickers and OSS Mix of tokenmaxxing/cost fatigue, nemotron coalition,…

model-releasesclem-delangue--x

There was an inflection point recently where the tide shifted to model pickers and OSS Mix of tokenmaxxing/cost fatigue, nemotron coalition, harness step functions, brains/claws, etc Long live those w

Wasn't Krea 2 supposed to be released ?

model-releasesr-stablediffusion

Krea 2, Krea's first foundation image model built from scratch, was announced on May 12, 2026 , with a focus on aesthetics, style transfer, and creative control . Krea 2 became available to everyone s

Wordle 1,813 4/6 ⬛⬛⬛🟨⬛ ⬛⬛⬛⬛🟨 🟨🟨🟨⬛⬛ 🟩🟩🟩🟩🟩

model-releasesanthropic--x

This post documents a Wordle game result shared by Anthropic on X (Twitter), showing puzzle #1,813 solved in 4 attempts with a final correct answer displayed through colored emoji tiles (green indicat