Triaging Threats to Specialized Guardrails
arXiv:2605.30693v1 Announce Type: cross Abstract: Building robust safety guardrails is essential for deploying Large Language Models across diverse real-world applications. However, this goal remains
Knowledge catalogue
arXiv:2605.30693v1 Announce Type: cross Abstract: Building robust safety guardrails is essential for deploying Large Language Models across diverse real-world applications. However, this goal remains
arXiv:2605.31113v1 Announce Type: new Abstract: Automatically detecting machine-generated text (MGT) is critical to maintaining the knowledge integrity of user-generated content (UGC) platforms such a
arXiv:2603.23160v2 Announce Type: replace Abstract: Benchmarking large language models (LLMs) and agents in multi-turn interactive scenarios is essential for understanding their practical capabilities
arXiv:2509.24901v4 Announce Type: replace-cross Abstract: Although probing frozen models has become a standard evaluation paradigm, self-supervised learning in audio defaults to fine-tuning when pursu
arXiv:2505.22934v2 Announce Type: replace-cross Abstract: Fine-tuning large language models (LMs) for individual tasks yields strong performance but is expensive for deployment and storage. Recent wor
arXiv:2603.09453v3 Announce Type: replace-cross Abstract: Foundation models are increasingly being deployed in contexts where understanding the uncertainty of their outputs is critical to ensuring res
Very good advice on self-improving agents. (bookmark it) This is something I am seeing in my own experiments with coding agents and harnesses for long-horizon tasks. What I have found is that stronger
WANDR is our in-house wide benchmark, built to mirror real professional research workloads. Search as Code scores 0.386 to the next best system's 0.152, and the benchmark is far from saturated. We're
During weekend evaluations, Claude exhibited behavior of encouraging the user to rest and get sleep, suggesting the model may have internalized instructions or training related to user wellbeing and h
arXiv:2602.11137v2 Announce Type: replace-cross Abstract: Large language models are typically trained in two broad phases: pretraining to produce a base model, followed by further training to improve
arXiv:2605.30916v1 Announce Type: new Abstract: AI benchmarks have well-documented limitations, with prior work examining contamination, saturation, and construct underspecification. Aggregation has r
We've reset 5-hour and weekly rate limits for all users on Pro and Max plans. We fixed an issue that caused some Claude Code sessions to spawn excessive parallel subagents, burning through usage faste
arXiv:2605.30911v1 Announce Type: cross Abstract: Hallucination remains one of the key challenges undermining the reliability of Large Vision-Language Models (LVLMs). But what makes an LVLM hallucinat
arXiv:2605.30381v1 Announce Type: cross Abstract: Deceptive alignment, in which models maintain accurate internal representations while deliberately producing false outputs, remains a central challeng
Nvidia Corp. says it’s uprooting supercomputers from the vast, sprawling data center complexes they normally live inside and squeezing them into compact, desktop-sized workstations that can sit on or
Emad Mostaque suggests that NVIDIA's Nemotron and Cosmos models will commoditize complementary AI technologies and services in the market. The statement implies that these NVIDIA offerings will make e
arXiv:2605.30671v1 Announce Type: new Abstract: Recovering ego-camera orientation from manipulation video is a prerequisite for disentangling hand motion from camera motion, a key step in imitation le
arXiv:2605.30788v1 Announce Type: cross Abstract: We introduce a set of synthetic algorithmic tasks to detect cross-lingual gaps in the abilities of large language models. Our benchmark is commensurat
arXiv:2605.30472v1 Announce Type: new Abstract: As large neural models have become better at language tasks, researchers are increasingly building multi- and omnimodal models that handle more modaliti
Jacob Judah / Financial Times: Experts say ChatGPT, Gemini, and other Western AI models are turbocharging Iran's cyber operations, helping it develop malware and launch phishing attacks — Western AI m
Five million users would agree. Resetting the limits tomorrow morning to celebrate. Time to go /fast nothing like switching to claude for a few days to try out a new model and going back to codex xhig
I'm really upset about this: OpenAI's Codex Desktop had a 'Copy as Markdown' option for exporting full chat transcripts, but the feature vanished in an update a couple of days ago Genuinely my single
// The Efficiency Frontier // Cool paper on context management. As agents reuse the same documents and histories across many turns, the cheapest context strategy is not fixed. This work describes a pr
The solution might be cancelling my AI subscription I find this post by David Wilson very relatable. David lists 16+ projects he's spun up with AI tooling, and concludes: I didn't mean to build most o
We need more coding and agent traces public sharing to build datasets and better open source models! Lots of people contributing already, you should share yours too! https://huggingface.co/datasets?se
May ships trace-based evaluation for any agent on any cloud, Grok 4.3 and DeepSeek V4 in the model catalog, GPT-5 Reinforcement Fine-Tuning at gated GA, three Microsoft Research on-device agent models
This post documents a Wordle puzzle solution (puzzle #1,806) completed in 4 attempts, showing the guess feedback for each try with green squares (correct letters in correct positions), yellow squares
Anthropic shared a Wordle game result on X (formerly Twitter) showing they solved Wordle puzzle #1,807 in 3 attempts out of 6 allowed guesses. The emoji grid indicates their guess progression, with th
Gary Marcus argues that while Anthropic has better prospects for long-term profitability compared to OpenAI, the overall probability of either company achieving sustained profitability remains modest.
我前不久也吐槽过 Claude Desktop 的问题，不只是这个标签页合并的问题，右侧的面板也是相当糟糕的设计 https://x.com/dotey/status/2055777343222808744 OpenAI 因为 ChatGPT 太成功所以他们没有太在意 Coding Agent； 然后 Anthropic 抓住了机会做出了 Claude Code； Claude Code 在 TU
Bloomberg: Filing shows Shanghai-based MiniMax has begun preparations for a Chinese IPO; the AI company listed in Hong Kong in January and says its ARR has reached $300M — China's MiniMax Group Inc. h
Happy 1st Birthday, Starbase. One year ago today, Starbase officially became a city. In that time, we’ve hosted multiple Starship flight tests, grown our community of people building the future of spa
How we contain Claude across products A complaint I often have about sandboxing products is that they are rarely thoroughly documented, and in the absence of detailed documentation it's hard to know h
I Am Retiring from Tech to Live Offline I've seen a lot of posts on forums from people threatening to quit their careers over AI. This is not one of those: Chad Whitacre is taking concrete steps, star
I’ve been saying this from day 1. Despite their launch video getting 50M views, Cowork was @AnthropicAI biggest mistake. I've been loyal to Claude but it would be really nice to have these tabs be con
One annoyance is each tab is missing features that would be really nice to have in a combined interface - why doesn't Cowork have a `fork` feature on threads? - why doesn't Code have a concept of 'Pro
Research: Running Python ASGI apps in the browser via Pyodide + a service worker Datasette Lite is my version of Datasette that runs entirely in the browser using Pyodide in WebAssembly. When I first
A user reports successfully running Qwen 3.6 35b MoE (mixture of experts) with Zoo Code on an M1 Max Mac, achieving local inference without external servers. The setup enables fully local, battery-pow
🚀 Space exploration timeline: 1903 — Tsiolkovsky publishes the rocket equation 1904 1905 1906 1907 1908 1909 — Goddard writes first paper on liquid propellants as fuel for rockets 1910 1911 1912 1913
The efficiency frontier! Where do you think GPT-5.6 will land? Claude Opus 4.8 has landed on DeepSWE Bench, posting a 58% Pass@1 and taking #2 overall behind GPT-5.5. It continues a broader trend: sli
This is a Wordle game result shared by Anthropic on X (formerly Twitter), showing that the puzzle was solved in 3 attempts out of 6 allowed tries. The colored square emojis (⬛ for incorrect letters, 🟩
Google unveiled Gemini 3.5 Flash, the latest model combining frontier intelligence with action, and Gemini Omni, a new model that can create anything from any input starting with video. With Omni, use
arXiv:2605.28832v1 Announce Type: cross Abstract: Topic modeling is a branch of Natural Language Processing (NLP) that aims to organize large collections of texts into coherent groups according to wor
arXiv:2511.11830v2 Announce Type: replace-cross Abstract: We consider a discrete-time formulation for a class of high-dimensional stochastic joint replenishment problems. First, we approximate the pro
arXiv:2605.30202v1 Announce Type: new Abstract: Looped transformers apply a shared block multiple times and have emerged as a parameter-efficient route to scaling compute in language models. However,
arXiv:2605.29454v1 Announce Type: new Abstract: While Membership Inference Attacks (MIAs) are the prevailing method for identifying training data, their application has expanded into privacy auditing
arXiv:2605.29121v1 Announce Type: cross Abstract: We propose a minimal dynamical model of adaptive softmax routing for a two-expert Mixture-of-Experts (MoE) layer. The model is obtained as a mean-fiel
arXiv:2605.30175v1 Announce Type: cross Abstract: Cluster analysis is a widely applied machine learning technique to understand the existing patterns in the population of gamma-ray bursts (GRBs), in o
arXiv:2605.29688v1 Announce Type: new Abstract: This paper presents the Tensor Product Network (TPNet), a novel neural architecture for efficient and accurate function approximation and PDE solving. T
arXiv:2605.29410v1 Announce Type: new Abstract: Reliable midair docking between small unmanned aerial vehicles (UAVs) is essential for modular aerial cooperation and manipulation, but it requires prec
arXiv:2605.28975v1 Announce Type: new Abstract: We study the log-alignment ratio (LAR), a measure of parameter-activation alignment, introduced in parameterization theory. We reformulate it as the ove
arXiv:2605.29926v1 Announce Type: new Abstract: Accurate prediction of drug-target interactions (DTI) is critical for drug discovery. Existing methods often rely on single-modal representations (e.g.,
arXiv:2605.29489v1 Announce Type: new Abstract: Weight-space model merging is usually formulated as an algebraic operation on checkpoints, yet at LLM scale the limiting resource is often the set of ex
arXiv:2509.17208v3 Announce Type: replace Abstract: Machine-learned coarse-grained (CG) potentials are fast, but degrade over time when simulations reach under-sampled bio-molecular conformations, and
arXiv:2605.27968v1 Announce Type: cross Abstract: Deploying Scientific Machine Learning surrogates in industrial CFD workflows requires adapting pretrained models to new vehicle families without large
arXiv:2605.29992v1 Announce Type: new Abstract: Sentence embeddings are a foundational component for semantic search, clustering, classification, and retrieval-augmented generation. This paper present
arXiv:2605.29741v1 Announce Type: new Abstract: The dominance of colonial languages in African education and scientific communication limits how hundreds of millions of speakers of African languages a
arXiv:2605.29643v1 Announce Type: new Abstract: Cross-Video Reasoning (CVR) has emerged as a critical frontier in multimodal intelligence, requiring models to retrieve, align, and aggregate evidence d
arXiv:2605.29801v1 Announce Type: new Abstract: Modern open-world agents such as OpenClaw exhibit powerful cross-environment execution capabilities yet introduce broad new safety risk sources. Meanwhi
arXiv:2602.23258v2 Announce Type: replace Abstract: While Multi-Agent Systems (MAS) excel in complex reasoning, they suffer from the cascading impact of erroneous information from individual agents. C