2026-09-24

Hey there folks! Aritra here from Hugging Face. I wanted to update you all about the latest changes in `transformers`. We now natively support GGUFs (llama cpp quants). You can use it like so: from transformers import AutoModelForCausalLM, AutoTokenizer model_id = "unsloth/Qwen3.5-4B-GGUF" filename = "Qwen3.5-4B-Q4_K_M.gguf" model = AutoModelForCausalLM.from_pretrained( model_id, gguf_file=filename, ) After loading, you're using the normal Transformers APIs. Why did we want to do this? Quantized

Disclaimer: I'm an engineer at Mulga, the company behind this. Self-promo, but it's AGPL-3.0 and free to run. The Problem It Solves Spinifex reimplements the AWS APIs on hardware you own: EC2, EBS, S3, VPC, IAM, ALB/NLB, EKS, ECS, ECR and RDS. It's real compute: instances are actual QEMU VMs, and VPCs are real OVN networks with security groups and elastic IPs. It's not an emulator like LocalStack. Why not existing FOSS Most infrastructure tooling, docs and skills today assume AWS. If you want th

It's unclear what will happen if FBI misses ShinyHunters' deadline.

Financial extortion doesn't seem to be the group's goal this time.

Meta says it issued a fix for the Muse zero-day vulnerability that would have let attackers do “whatever” they wanted on a victim’s Mac, highlighting the inherent dangers of AI helpers.

Introducing private, server-side memory to Private AI Compute for personal AI.

arXiv:2609.25337v1 Announce Type: new Abstract: Dialogue failures in language models are usually framed as memory failures: context too long, summaries lossy, a constraint forgotten. We argue this misses a deeper problem: in many conversations the model does not forget, it commits too early. An ambiguous early turn collapses into a single hidden interpretation, and later clarification is filtered through that commitment. We call this early posterior collapse: unresolved user intent collapsing in

overshadowing more efficient GPT6 models from OpenAI

arXiv:2609.25199v1 Announce Type: new Abstract: Lean Pool is a repository of formalized mathematics. It is grown, maintained and optimized by AI agents.

arXiv:2609.25286v1 Announce Type: new Abstract: Structured-data agents in enterprise settings must reason over complex data environments whose relevant evidence is distributed across schemas, relationships, policies, and recurring business roles. Modern agentic systems often address this burden through reusable markdown-style memory or skill files that preserve previously discovered information for later queries, reducing the need to rediscover the same structure repeatedly. This is useful, but

arXiv:2609.25285v1 Announce Type: new Abstract: Finding circuits in language models usually means running many careful interventions. We try something simpler: treat attention as a routing map from one forward pass, keep a small set of routes that point toward the answer, and ask whether those routes actually matter. They often do. On induction and IOI (tasks where the "right" circuit is already known), ablating our extracted edges hurts the model much more than ablating a random set of the same

LLM-as-a-judge enables evaluation across diverse tasks, but inference cost and confidence reliability become critical at scale. We study whether a decision-only judge can provide an economical first pass and identify when stronger evaluation is needed. Comparing jev-as-a-judge with sixteen generative and reward-model judges, with blinded human adjudication, we find it within three percentage points of a state-of-the-art LLM judge, our strongest comparator, on ordinary preference and evidence-gro

GPT-6 Astra produces more structured, context-aware legal documents, freeing lawyers to focus on strategy.

Hey HN! We’re Vincent and Jochen from Sitefire ( ). We have been working together for years, with backgrounds in RL/optimization at Stanford and software engineering from Technical University Munich (TUM). With Sitefire (YC W26), we help marketing teams get recommended by AI Search (ChatGPT, Google AI Overviews, AI Mode, Claude, etc.). Our software monitors prompts, sees which web pages get cited, and uses these insights to help marketing teams take action, e.g. create YouTube videos or write th

Article URL: Comments URL: Points: 63 # Comments: 12

9月23日云栖大会期间，斑马智能发布新一代全模态端侧大模型AutoOmni 2.0-23B-A3B

In this report, we introduce Ovis-Embedding, a state-of-the-art omni-modal embedding family built on native integration of text, image, video, and audio. Instead of assembling separate modality towers, Ovis-Embedding uses a shared multimodal backbone to encode different modalities in a common representation space. Specifically, we make three key advances: (1) native omni-modal initialization: we adopt a pretrained Qwen-omni model as the embedding backbone and adapt it through contrastive trainin

Kyutai has released Voice of Reason, 2 open-weight speech-to-speech models built on GLM-4-Voice-9B. Supervised fine-tuning and reinforcement learning lift spoken GSM8K accuracy from 27.3% to 77.1%. There is no transcription step and no text LLM in the loop. Both checkpoints are on Hugging Face and run on a single H100. The post Kyutai Releases Voice of Reason: A Speech-Native Model that Solves Spoken Math with Reinforcement Learning appeared first on MarkTechPost .

SF October 14th: A Birds of a Feather Session on Agentic Engineering I'm hosting an evening event with Jesse Vincent in San Francisco on Wednesday 14th October for people who are building weird and interesting things with and on top of coding agents. Think of it as an agentic show-and-tell: ​Compare notes with other builders and experimenters on things you’re trying, what you're learning, and what you haven’t figured out yet. We’re especially interested in work you haven’t discussed publicly, od

The rising global prevalence of mental health conditions, together with longstanding barriers in traditional healthcare, such as limited resources, high cost, stigma, and privacy concerns, has created an urgent need for accessible and scalable support. Large Language Models (LLMs) have emerged as a transformative technology with strong potential to democratize mental health support through advanced natural language understanding and generation. However, the rapidly expanding, fragmented body of

Learn how Airbnb is expanding access to GPT-6 Astra and OpenAI frontier models to help engineering teams solve bugs, design systems, and ship faster.

The rumor about Kimi execs getting arrested finally has some legs. I believe the reality is more like under investigation for potential arrests or fine.

Update from : It's been about a week so I'm back. People were asking me about the model. People wanted code, or models, etc. Most of that is useless to you right now because you're not going to use an under trained model. So let's get to the details. The spec locked to the following after a LOT of testing : 2.6b model all up. Embedding, LM head, AttnRes, etc. 2.2b are trained. Embedding / LM Head are frozen (~205m each) 4.3b ENGRAM table. Yes. She's chonky. Architecturally speaking now : This in

I hear Qwen code unlocks the model better. I also think it has more power user features than open code? It’s nice open code can work with multiple models easier though Thoughts?

Trump focus on winning “AI race” may deter China from sharing safety intel.

We present a compact geometry-native latent space as a shared foundation for perception and generation. Visual generators can produce photorealistic frames without preserving a consistent 3D scene. We argue that this is not only a modeling problem but also a representation problem: generators typically evolve appearance-centric latents, while perception models recover geometry in a semantically rich space that encodes cross-view structure. Rather than adding geometry as another output, we repara

As the US and China race to become the dominant power in the AI industry, the countries also appear to be figuring out ways to communicate on national security issues.

Multilingual scene text recognition (STR) remains challenging due to the scarcity of training data for most languages and the difficulty of serving diverse scripts within a single model. Existing solutions either deploy one recognizer per language, inflating cost and introducing error accumulation, or rely on massive vision-language models (VLMs) that are expensive and still inaccurate on many scripts. In this work, we pursue an all-in-one multilingual recognizer that is simpler than per-languag

As concern over AI safety and rogue agents continue to make headlines, it’s no surprise that cybersecurity stocks are rising, or that investors are pouring massive amounts of capital into startups trying to build the next generation of security for an AI-native world. We’re even seeing companies like Instinct and Simile bring in nine-figure checks and valuations that wouldn’t have made sense a […]

Tool: Shadow roots, explained with live examples Prompt to Fable 5.1 Medium: Build an artifact to explain shadow roots in CSS with interactive examples Tags: css

OpenAI is extending access to its Daybreak program to the Government of Ukraine to support the cyber defense of civilian infrastructure.

Using GPT-5.6, Ringg powers multilingual agents across voice, chat, WhatsApp, and web for 90% less cost vs. GPT-4.1.

Radical Numerics is using biological chain-of-thought and multimodal perception to keep up with the bio-defense arms race, design new genomes and gain insights into biology itself.

The Biological Computing Company is bringing its AI tools to Amazon Web Services in a major boost for a once-fringe field that aims to marry nature with code.

Language models can be adapted by changing the computations applied to individual tokens. Quantum circuits offer one such approach, but evaluating wider circuits inside a large model can be computationally demanding. Here we introduce HyperQ, which adds token-conditioned quantum residual branches to a frozen masked-diffusion language model. A quantum residual branch is a module in each transformer block that reads a token's hidden state, emits the coordinates of that token's circuit, executes it

arXiv:2609.25254v1 Announce Type: new Abstract: Analyzing neuroimaging data requires specialized coding and statistical expertise, which limits accessibility for researchers without computational backgrounds. We present the AI Neuroscientist, a language agent for interactive data exploration. The system integrates a large language model (LLM) with a neuroimaging toolset to perform quality control, modeling, and visualization. This allows researchers to query data quality and specify analysis par

arXiv:2609.25187v1 Announce Type: new Abstract: Task planning bridges high-level instructions and executable behavior in long-horizon manipulation, yet modern Vision-Language-Action (VLA) systems often leave this intermediate structure implicit. Existing chain-of-thought (CoT) planners also tend to rely on coarse task-level annotations or serialize long reasoning traces token by token. We present X-Planner, a planning front-end that addresses both the supervision and representation of embodied r

Vector Quantization (VQ) is fundamental to discrete visual tokenizers that power modern autoregressive and masked image generation models. While recent shared-projection codebook methods have substantially advanced codebook utilization, training stability remains a critical and underexplored challenge. We argue that the root cause lies in the entanglement of the Encoder--Decoder and Codebook training: because neither module can reliably fulfill its own responsibility in isolation, the system can

50% per quarter is amazing. 4× faster than DNA sequencing, 6× faster than compute, 18× faster than lithium batteries, and (up to 1973) 54× faster than electricity. Every year moving forward is going to be significantly different that the prior year. What do you think? We will be running coding agents on our phones pretty soon.

Hey everyone, This is a worn out topic, but I still want to bring it up once again because I keep finding it really difficult to talk about IRL with people close to me professionally without having the discussion go off-rails. I have this feeling that people have quite strong almost dogmatic views on this in either direction and it's really throwing me off. For some background: I'm a mid level engineer currently working at a mid-sized company. This far in my career I've been viewed as a quite st