All entries

What I like about Jev: for years, large generalist models sucked almost all the oxygen out of AI. But there are huge opportunities in much more specialized and customized models, built for specific ta

arXiv:2609.21523v1 Announce Type: new Abstract: A system may be compressed before its downstream task is fully known. We ask how much retained state is then necessary and how much can be saved by limi

arXiv:2609.21924v1 Announce Type: new Abstract: Interactive retrieval under partial evidence is a sequential information-acquisition problem: an agent must decide which question will create the most u

arXiv:2609.20942v1 Announce Type: new Abstract: Large language models (LLMs) increasingly participate in scientific evaluation, both as automated reviewers and as assistants to human reviewers. As mod

arXiv:2609.21187v1 Announce Type: new Abstract: Agent models are frequently evaluated one decision at a time, where the model predicts the next action based on the gold interaction history, which is s

arXiv:2609.21247v1 Announce Type: new Abstract: Large Reasoning Models increasingly use intermediate traces for machine translation, but it remains unclear when such reasoning helps or hurts. We analy

arXiv:2609.21412v1 Announce Type: new Abstract: Medical image segmenters often get worse when sites, scanner vendors, or protocols change. Continual test-time adaptation (CTTA) addresses this problem

arXiv:2609.21942v1 Announce Type: cross Abstract: A robot that fails at a task faces the first decision in corrective dialogue: act on its own diagnosis, consult another onboard sensor, or interrupt a

arXiv:2609.21734v1 Announce Type: new Abstract: Designing effective Human-Robot Interaction in task-oriented settings requires carefully balancing user engagement with socially acceptable levels of ro

arXiv:2609.21662v1 Announce Type: cross Abstract: Activation steering has become a widely used approach for controlling language models during explicit chain-of-thought (CoT) reasoning, motivating its

arXiv:2609.21185v1 Announce Type: new Abstract: Low-friction surfaces challenge bipedal locomotion by limiting the contact forces available during stepping. Inspired by penguin waddling, we investigat

I have been reasonably satisfied with my single R9700 (32GB) as I can run practical quants of Qwen 3.8-27B at good speeds, as well as other similar models in its weight class (Gemma 4 is still my go-t

where it stops: geometry. relative size 0.52, diagonal direction 0.65, while GPT-5.6 and Gemini 3.1 Pro get 100%. UI screens: finding what to click is ~100% for everyone, one step of reasoning 0.90 vs

Physical AI is moving rapidly from research to large-scale deployment. By 2035, ABI Research projects an installed base of 49 million level 3-5 autonomous vehicles (AVs), while Omdia estimates that ro

with jev, everyone is understanding the importance of calibrated confidence scores for discrete decision making we've taken that approach one-step further and created grounded confidence scores for ge

arXiv:2609.20892v1 Announce Type: cross Abstract: Closed-loop visual servoing requires predictions that indicate whether an action reduces task error, not only whether the action is plausible. We call

arXiv:2609.21748v1 Announce Type: new Abstract: Behavioral failures can make a transformer appear to lack a world model even when it has learned faithful representations of its environment. We demonst

arXiv:2609.21391v1 Announce Type: cross Abstract: Neural Radiance Fields (NeRF) have attracted extensive attention in recent years due to their strong capability for high-quality 3D reconstruction and

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 b

arXiv:2609.21770v1 Announce Type: new Abstract: Image registration is a vital preprocessing step in multimodal perception tasks, including image fusion, object detection, and semantic segmentation. In

Xiaomi: Xiaomi debuts open-weight omnimodal models MiMo-V2.6 Pro and Flash; Pro allegedly performs “on par with Opus 5 and GPT-5.6 Sol across most agent benchmarks” — Frontier intelligence, all the mo

https://huggingface.co/yandex/AliceAI-Foundation-80B-A3B-Base It's not a Qwen3 finetune, it's actually its own fully custom architecture. No Llama.cpp support yet sadly (Also note that this model is N

You can simply run any GGUF with llama.cpp with n_predict=1 and n_probs=10, disable reasoning, and prompt it such as 'If the following email is spam, respond with 1, if not spam, respond with 0. Do no

ZCode is now open source, and the reported security issues have been addressed. Source code: https://github.com/zai-org/ZCode The repo includes its desktop app, web workspace, backend, Agent CLI, and

arXiv:2609.21726v1 Announce Type: new Abstract: Reliable robotic grasping benefits from estimating the evolving physical interaction and selecting a grasp-dependent compression target. Tactile sensors

arXiv:2609.21712v1 Announce Type: new Abstract: Generative world models offer controllable and repeatable closed-loop simulation for end-to-end and vision-language-action driving policies, but product

5x faster than LLMs while being 13% cheaper than GPT 5.6 Luna, 88% cheaper than GPT 5.6 terra and 99% cheaper than Sonnet 4.6 is insane Has anyone else used Jev for evals? We tested Jev against LLM ju

This weekend, I did a fun experiment to create something similar to Jev. I LoRA fine-tuned Qwen3.5 4B using a mix of publicly available datasets and synthetic data. For the synthetic data, I used Deep

Jev from TypeSafe AI got me thinking about non-autoregressive decision engines, so I built a small proof-of-concept called AES to test the concept locally on open weights. It uses Qwen/Qwen2.5-1.5B-In

AGI FTW @ FT “The most popular AI models from ChatGPT, Claude, Copilot, Grok and Gemini provided wrong answers to financial queries 57 per cent of the time on average, according to research from techn

Owen Walker / Financial Times: AI companies, including OpenAI and Anthropic, are putting pressure on office rents in Singapore as they embark on expansion in response to government overtures — Anthrop

Hi all! i am currently working as a software engineer for a pretty big enterprise fintech company here in the states. In the the last 12 months at my job there has been a huge push for developers to u

Not a pic sharing thread, bunch of those already - mainly just a discussion about first impressions Very underwhelming to me personally. I know its a matter of settings but normally comfy default temp

OpenAI’s decision to retire custom GPTs appears to rest on the assumption that Plugins, Skills, Projects, Apps, and Workspace Agents collectively replace their functionality. For many Small and Medium

You can now automate EDD (eval-driven development) with Coding Harness Agents > build a baseline LLM-based application > score every change with judge evals > keep what improves, reject what regresses

chat : add dedicated Ling 3.0 (Bailing V3) parser (#28682) chat: add dedicated Ling 3.0 (Bailing V3) parser Ling 3.0 Flash templates pre-open the think block in the generation prompt, so the model nev

metal: add F16 input to the FWHT (#29094) metal: add F16 input to the FWHT The Metal FWHT kernel accepts F32 input only. This change makes the source type a template parameter, so the kernel reads an

fix(mamba) : make time-step projection input contiguous (#28832) mamba : make time-step projection input contiguous Assisted-by: ChatGPT mamba : skip contiguous copy after normalization Assisted-by: C

CUDA: enable sparse fa for qwen4 (#28770) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/48739244 macOS/iOS: macOS Apple Silicon (arm64) macOS Apple Silico

common/peg : handle invalid utf-8 sequences in the AST (#29161) common/peg : handle invalid utf-8 sequences in the AST cont : return maximal subpart per Unicode recommendations cont : remove strict ar

metal : support arbitrary hc in dsv4_hc_pre (#29169) the dsv4_hc_pre kernels hardcoded hc = 4 via a constexpr used with simd_shuffle, so the op was rejected by supports_op for any other hc and fell ba

CUDA: tune FA for Gemma 4 on Ampere or newer (#29152) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/48802880 macOS/iOS: macOS Apple Silicon (arm64) macOS

breaking: author of the study that so many “AI influencers” are taking as evidence that LLMs feel pain tells me “we don’t claim they feel pain.” The good news is that the author understands correctly

Claude not having an image generator is a gap in agentic projects that involve knowledge work. It is quite good at using code to draw or model objects, but image generators from Google and OpenAI give

Correction: “highly profitable” for Jensen, not for OpenAI or Anthropic or (afaik) their corporate customers. BREAKING: Nvidia CEO Jensen Huang says AI reached a major turning point this year, becomin

TV for the developer internet GitHub, Hacker News, dev.to, Hugging Face, and release notes, presented as channels you tune into, each auto-cycling its own stories, instead of five tabs to remember to