Knowledge catalogue
arXiv:2607.11920v1 Announce Type: cross Abstract: Evaluating decisions made under uncertainty is hard when labeled outcomes are scarce, costly, or confounded with luck. We treat subjective expected ut
someone just told me about this take* on CUA this is one of those gell mann moments for me lol. i've been watching computer use since World of Bits (Shi, fan, karpathy, hernandez & liang 2017). we wer
arXiv:2607.12161v1 Announce Type: new Abstract: Context-reduction layers for API-based coding agents, including command-output compressors, retrieval rankers, and payload-optimizing proxies, are usual
New model for AMD Strix Halo users: My 198B Step 3.7 Flash release was a big hit, but this one may be even better: 298B-parameter Hy3, now running on a new 2-bit FPX codebook designed to map efficient
Another big reason to use combination of frontier models. Chain-of-thought monitoring is treated as a reliable safety layer for agents. This DeepMind-affiliated study shows the layer can be argued out
ChatGPT still has study mode, but rather than /study you now have to type @ study It makes the AI act more like a tutor than a helpful assistant, and some work suggests it is better if you are trying
arXiv:2607.07721v1 Announce Type: new Abstract: Retrieval-Augmented Generation (RAG) and agentic frameworks have advanced enterprise AI considerably, yet agents remain fundamentally reactive: they wai
GROK 4.5 LEADS ON REAL PROFESSIONAL WORK BENCHMARK New data from Snorkel shows Grok 4.5 outperforming other frontier models on real-world professional tasks. On their GDPval+ benchmark (expert-created
arXiv:2607.08066v1 Announce Type: new Abstract: Chain-of-thought (CoT) monitoring is a promising safety mechanism for AI agents, based on the premise that visible reasoning traces can surface misalign
arXiv:2607.08065v1 Announce Type: new Abstract: LLM-as-judge (Zheng et al., 2023) is increasingly the default for evaluating AI systems in enterprise pipelines, often scaled to ensembles (Verga et al.
arXiv:2607.07707v1 Announce Type: cross Abstract: Limited memory language models (LMLMs) externalize factual knowledge during pretraining to a knowledge base (KB), rather than memorizing it in their w
On June 16th, Z.ai released GLM 5.2, its latest flagship model. At the time of announcement, it advertised scores at or near Anthropic and OpenAI's models, and far ahead of GLM 5.1. In the world of us
arXiv:2607.06597v1 Announce Type: cross Abstract: Public chest-radiograph (CXR) datasets are typically released with small, fixed label schemas such as CheXpert-14. However, the underlying free-text r
Many of the most challenging and valuable problems in the world are related to optimization. Now, AI is now making these problems tractable. If you've ever tried to design a microchip, plan a delivery
arXiv:2607.06636v1 Announce Type: cross Abstract: Large language models frequently generate code that appears correct on typical inputs yet fails on edge cases, invalid inputs, and other specification
// The Harness Effect // (bookmark it) Now more that ever pay very close attention to the orchestration harness and its effect on costs and performance. This study ran 22 evaluation tasks on six found
arXiv:2607.06906v1 Announce Type: new Abstract: Agentic AI development today runs on token maxing: buying capability with tokens -- longer reasoning traces, more turns, wider tool payloads, bigger rep
Perplexity conducted a cost efficiency comparison of language models, using GLM 5.2 as the baseline metric for cost per task. Results showed GLM 5.2 with an advisor achieved 2.1x cost efficiency on th
arXiv:2607.06799v1 Announce Type: cross Abstract: Evaluating uncertainty in AI-generated SQL queries requires estimating whether a query is correct, where correct means it executes to the same result
arXiv:2607.06413v1 Announce Type: cross Abstract: Large language model coding agents increasingly perform open-ended data modeling and analysis. These agents are stochastic and adaptive, and therefore
arXiv:2607.05614v1 Announce Type: cross Abstract: Document comprehension is a challenging yet impactful task for Multimodal Large Language Models, especially as these systems see growing adoption in r
arXiv:2510.19771v4 Announce Type: replace Abstract: LLM-based agents are increasingly moving towards proactivity: rather than awaiting instruction, they exercise agency to anticipate user needs and so
Didn't even have time to write up thoughts on my early access to the new GPT voice, but it is really good & much closer to the science fiction experience of talking to AI (in part because it is much s
Finally can share that I have been testing GPT-5.6 early and oooo boy. It is an execution beast. So much so that I think 5.6 is the absolute wrong name considering how big of a leap this felt to me. T
Grok 4.5 true usefulness is excellent We obtained early access to evaluate @SpaceXAI's newest Grok 4.5 model on GDPval+: real professional deliverables (from a variety of economic sectors and occupati
My team talks directly to *my* AI workforce. They can skip me entirely. It's the first time working with AI agents actually feels team-oriented and collaborative (and not just one person becoming more
Not another demo or benchmark. @ShoucongChen is a senior member of our technical staff. A real project, scoped at 1-month. Delivered in 4 days with GLM5.2 Fast. The best devs deserve >400 t/sec. Take
super interesting - and a reminder that language models model language and not ideas per se. also a stark example of how easily LLMs can (as a consequence) be influenced by propaganda. Asking ChatGPT
The anthropomorphization of this one is off the charts. Why does @AnthropicAI do this? It is intellectually lazy. Unscientific. New Anthropic research: A global workspace in language models. Of everyt
arXiv:2607.05552v1 Announce Type: cross Abstract: Large language models (LLMs) increasingly issue judgments read as binary verdicts, and a growing literature reports such judgments shifting under logi
WOW! GROK 4.5 PRICES ARE NEAR OPEN SOURCE HOSTED PRICES! This will move many that have token cost fatigue in corporations. But that is only half the story: it is ~4.2 more efficient using far less tim
Xiaomi now processes more AI tokens than OpenAI. On OpenRouter, Chinese models just crossed 45% of all token volume. Anthropic is at 15.3%. OpenAI is at 7.4%. For now, the frontier is American models,
arXiv:2607.04072v1 Announce Type: cross Abstract: Large language models can generate plausible quantum code, but it is unclear whether they can reliably target the specific software development kit (S
arXiv:2604.00392v2 Announce Type: replace-cross Abstract: Agents that synthesize their own tools ship a second artifact alongside each answer: a software library that future tasks reuse, compose, and
arXiv:2607.02598v1 Announce Type: new Abstract: Autonomous computational pathology (ACP) converts high-level pathology analysis goals into executable, traceable and clinically bounded workflows. Reali
Excellent investigation of the mechanics of language models. 🫢A notable blindspot, however: several teams have actually been **directly** comparing the working of LLMs to those of the human brain 🧠 fo
arXiv:2607.04631v1 Announce Type: new Abstract: The cost of producing code is rapidly diminishing with increasingly capable AI agents, while quality assurance of generated programs has not kept pace.
arXiv:2607.03935v1 Announce Type: new Abstract: Self-evolving frameworks usually optimize task solutions while treating the surrounding harness as fixed. We introduce Harness-Aware Self-Evolving (HASE
I have started including a very strange line in my Fable prompts. And I think it's working. It's some version of: 'I know you're incredibly goal-oriented, and I want to structure this around goals. Fi
I wonder how this could be used to study creative thinking and the generation of new ideas. If I understand correctly, the J-space is some sort of internal workspace where the model holds concepts to
imo this is the most impt part of anthropic's J-space paper today. it's a two-parter: 1) ant proved that they can do 'brain surgery' interventions into reasoning to change topics midstream* 2) THE MOD
arXiv:2511.17384v2 Announce Type: replace-cross Abstract: While Visual Large Language Models (VLLMs) show great promise as embodied agents, they continue to face substantial challenges in spatial reas
arXiv:2607.02825v1 Announce Type: cross Abstract: We release extsc{JavaVulBench}, a benchmark dataset and evaluation harness for Java vulnerability detection. The dataset contains sim30{,}600 Java met
arXiv:2607.05391v1 Announce Type: new Abstract: Scaling pre-training, post-training, and test-time compute have become the central paradigms for improving the capabilities of LLMs. In this work, we id
arXiv:2602.14095v2 Announce Type: replace Abstract: Monitoring chain-of-thought (CoT) reasoning is a foundational safety technique for large language model agents; however, this oversight is compromis
NEW AI paper worth bookmarking. This is something I called early, and this paper confirms it: verification has emerged as a new important scaling axis. Here is the simple explainer and what this paper
Please please don’t let the Chinese steal from us ,but can we please have a huge copyright exemption? 🙏 Big story here on Anthropic’s moves in Australia, pushing for huge data centre deals and a huge
arXiv:2607.04438v1 Announce Type: cross Abstract: Research dissemination, turning a paper into a poster, a talk video, and a blog post, is still a manual last mile. Prior automation treats each artifa
arXiv:2607.02802v1 Announce Type: cross Abstract: As LLMs are increasingly deployed as autonomous adjudicators in semi-open textual game environments, robust rule adherence becomes critical when user
arXiv:2607.03451v1 Announce Type: cross Abstract: While skill optimization for autonomous agents has gained traction, existing methods rely on complex pipelines. This leaves a fundamental question una
arXiv:2511.07403v2 Announce Type: replace-cross Abstract: Multimodal large language models (MLLMs) have achieved remarkable progress in vision-language tasks, but continue to struggle with spatial rea
The AI labs desperately need non-engineer Peters, Borises, and Thariqs. Most demos for 'business users' are about replying to emails or to Slack. And yes, that's helpful to manage the cacophonous hell
The problem with Anthropic's consciousness paper My last post got more attention than I expected, and the question I keep getting is some version of 'okay, so what is actually wrong with the paper?'.
arXiv:2603.18482v2 Announce Type: replace Abstract: Standard decoding strategies for text generation, including top-k, nucleus sampling, and contrastive search, select tokens based on likelihood, rest
This is a good clip that shows a few layouts. I'm pretty happy with this, maybe the slide layout could be a bit different and I'm curious how it will alternate between camera cuts. That's all for now,
arXiv:2603.17673v2 Announce Type: replace-cross Abstract: LLM agents are becoming increasingly important in the security domain, but leading systems are often closed-source, cloud-based, hard to repro
arXiv:2607.02931v1 Announce Type: new Abstract: AI tools are accelerating scientific publication while the systems that review it struggle to keep up, and independent verification of published researc
arXiv:2607.05189v1 Announce Type: cross Abstract: Persistent personal agents combine long-term memory with access to users' external environments, enabling personalized foreground assistance and proac
Interesting stuff. And the visualization at the end is worth trying: https://www.neuronpedia.org/qwen3.6-27b/jlens New Anthropic research: A global workspace in language models. Of everything happenin
One of the only times I remind people I have a PhD in computational neuroscience is when people without a neuroscience background say their model works 'like the brain.' In these cases, I put on my ne