model releases

TRACE: Temporal Audit and Condition-aware Evaluation of Streaming Video Understanding

model-releasesarxiv-cs-cl

arXiv:2609.30670v1 Announce Type: new Abstract: Streaming video understanding requires models to interpret evidence as it arrives, yet current evaluations often report task scores without specifying w

Human

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Model Releases

TrafficImag: A Benchmark for Counterfactual Roadside Traffic Video Generation

DGX agent

arXiv:2609.30722v1 Announce Type: cross Abstract: Existing roadside traffic datasets support perception, forecasting, and visual question answering, but they do not evaluate counterfactual video gener

model-releasesarxiv-cs-ai

28 Sept 2026

Trained locally: ultra-fast 0.8B/2B System 1 decision models that match Jev on benchmarks and Doom, ~30 ms per decision (open weights)

TL;DR: The Jeff models are a set of Qwen3.5 and Gemma fine-tunes for zero-shot classification: small, efficient, open-weight models with respectable out-of-the-box performance that can be slotted righ

Training-Free Bottleneck Width Planning for Convolutional Autoencoders

arXiv:2609.30755v1 Announce Type: new Abstract: Multiscale Spectral Rate-Distortion (MS-SRD) estimates the bottleneck channels required at user-supplied spatial cuts from training images and a normali

UltraG-Bench: A Multi-task Benchmark for assessing Large Vision-Language Models on Pixel-level Evidence Grounding in Ultrasound

arXiv:2609.30928v1 Announce Type: cross Abstract: Ultrasound is one of the most widely used medical imaging modalities, and recent large vision-language models(VLMs) have shown increasing capabilities

Uncertainty and Explainability in Deep Rough Volatility: A Neural Information-Theoretic Posterior Approach

model-releasesarxiv-cs-lg

arXiv:2609.31570v1 Announce Type: cross Abstract: Deep learning has substantially accelerated the calibration of complex stochastic-volatility models, but neural point calibration alone does not captu

Understanding Perturbed Parameter Ensemble Sensitivities Using A Contrastive Learning Approach

arXiv:2609.30420v1 Announce Type: cross Abstract: Perturbed parameter ensembles (PPEs) reveal how physics parameters affect climate simulations, but interpreting parameter sensitivities across multiva

Understanding the Role of Prompt Template in Knowledge Distillation for Safety Alignment

model-releasesarxiv-cs-cl

arXiv:2609.30802v1 Announce Type: new Abstract: Prior research has demonstrated that the choice of prompt template during Supervised Fine-Tuning (SFT) significantly impacts the robustness of safety al

Unknown-Traffic Detection, Calibration and Shortcut Reliance in Distilled Encrypted-Traffic Classifiers over One Year

model-releasesarxiv-cs-lg

arXiv:2609.31141v1 Announce Type: cross Abstract: Knowledge distillation is the standard way to compress encrypted-traffic classifiers for the edge, and almost all such work judges students by accurac

VFM-UDA++: Improving Network Architectures and Data Strategies for Unsupervised Domain Adaptive Semantic Segmentation

arXiv:2503.10685v3 Announce Type: replace Abstract: Unsupervised Domain Adaptation (UDA) enables strong generalization from a labeled source domain to an unlabeled target domain, often with limited da

ViSTA: A Simple Bridge Extends Visual Alignment to Clinical Time-Series Understanding in Multimodal LLMs

arXiv:2609.31448v1 Announce Type: cross Abstract: Clinical prediction models estimate risk from patient measurements, while large language models support medical text understanding and question answer

VLALight: Lightweight Vision-Language-Action Models for Emergency-Aware Traffic Signal Control

arXiv:2609.30709v1 Announce Type: cross Abstract: Traffic signal control (TSC) is essential for mitigating urban congestion. Recent advances in vision-language models (VLMs) enable richer interpretati

Watch me post-train AliceAI-Foundation-80B-A3B from base to instruct at home, live, on my V100s!

No click bait baby I promise - I'm live streaming the training process kinda like MiMo. Original thread: https://www.reddit.com/r/LocalLLaMA/comments/1wpg4a8/im_trying_to_posttrain/ If you didn't see

@WeAreDevs Here's how I define 'Fable class models' - first Claude Fable 5, now Claude Opus 5.5 and GPT-Astra 6 and maybe GPT-5.6 Sol as wel…

model-releasessimon-willison--x

Simon Willison defines “Fable class models” as a lineage of generative AI systems that began with Claude Fable 5, progressed to Claude Opus 5.5, and later included GPT‑Astra 6 (with a possible additio

WeaveAgent: A Two-Stage Tool-Routing Agent for Ultra-High-Resolution Remote Sensing Imagery

arXiv:2609.31234v1 Announce Type: new Abstract: Problem. Ultra-high-resolution (UHR) remote sensing with vague user intents has two bottlenecks: visual tokens are expensive, and tool calling must be f

What Improves Multimodal Misinformation Detection? Answers from a Large-Scale Empirical Study

arXiv:2609.30402v1 Announce Type: cross Abstract: Multimodal misinformation is increasingly crafted to look convincing by pairing a textual claim with an image that appears to 'prove' it. Yet in pract

When can we say AI made a scientific discovery?

model-releasesmit-tech-review

This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. Last Wednesday, Anthropic announced that earlier this year it

When the Preconditioning Exponent Turns Negative: Learning-Rate Coupling and Cross-Environment Generalization

model-releasesarxiv-cs-lg

arXiv:2609.30271v1 Announce Type: new Abstract: Adaptive optimizers are commonly parameterized by a fixed power of the second-moment estimate. Existing partially adaptive methods study exponents betwe

Where Compute Matters: Heterogeneous Attention for Efficient Video Diffusion

arXiv:2609.31050v1 Announce Type: new Abstract: Efficient video generation requires reducing the quadratic cost of self-attention over long spatio-temporal token sequences. Existing efficient-attentio

Which Influence Are We Estimating? The Role of Counterfactual Specifications in Data Attribution

arXiv:2609.31214v1 Announce Type: new Abstract: Estimating the influence of training examples on model behavior is essential for data debugging, valuation, and attribution. Existing influence estimato

While the token prices for Sonnet 5.5 remain the same as Sonnet 5, Anthropic reports more than 30% faster output generation and up to 30% lo…

model-releasescognition-ai--x

While the token prices for Sonnet 5.5 remain the same as Sonnet 5, Anthropic reports more than 30% faster output generation and up to 30% lower cost per task than Sonnet 5, with unchanged token pricin

who's got two thumbs and just ranked #6 of 49 on image-jevbench? 👍 this guy 👍 one of the fastest in top 10, accuracy is close. calibration…

model-releasesyohei-nakajima--x

who's got two thumbs and just ranked #6 of 49 on image-jevbench? 👍 this guy 👍 one of the fastest in top 10, accuracy is close. calibration of confidence is where it's weakest. @airesearch12 says a sim

Why your startup needs open models alongside frontier APIs

model-releasesgoogle-cloud-ai

Every week, I talk with founders who are building at an unbelievable pace. Teams are moving from inception to product-market fit faster than ever, with foundation models wired deeply into their core p

Words Speak Louder Than Order: A Behavioral Evaluation of Gemma 4

arXiv:2609.30716v1 Announce Type: cross Abstract: When a language model receives two conflicting documents as input, how does it decide which one to prioritize? Does it rely on how the sources are fra

worst these will ever be 3:26 section is a good example of super persuasion incoming right in the kokoro

model-releasesemad-mostaque--x

worst these will ever be 3:26 section is a good example of super persuasion incoming right in the kokoro Made with Claude Opus 5.5. The fearmongering about AI always makes us forget that NOTHING WENT

ZooWork-ShopRanker: An Open, Preference-Aligned E-Commerce Reranker

model-releasesarxiv-cs-cl

arXiv:2609.31002v1 Announce Type: new Abstract: Open rerankers trained for general web retrieval transfer imperfectly to e-commerce, where ranking decisions depend not only on topical relevance but al

2026 in LLMs (so far)

model-releasessimon-willison

On Friday I gave the closing keynote at the WeAreDevelopers World Congress North America in San Jose. I tied together the key trends from the past year into a chronological exploration of everything t

2x Tesla p100s, q6_k quant, Qwen 3.8 27B ~60tps V3.0

Hey guys! I have been excited to share this here. This is a project consisting of kernel optimizations for the Tesla p100 series graphics card ($80). I want to start by saying I am 17 years old and do

Adding logit penalty for 'wait', 'maybe' and 'perhaps' to Qwen models improves their accuracy

Meta came out with a banger paper https://arxiv.org/pdf/2606.00206, but it did not look at various quantizations supported in llama.cpp. So I did a run on 50 random MATH-500 questions (https://hugging

Advice on models for RAG use case

Hi all, I'm looking for some advice picking models. I'm looking to try a project to ingest quite a large amount of docments into a knowledgebase, allowing me and others to ask questions about the data

Also kudos to fireworks AI for admitting that they built on top of Kimi (I remember the awkward Mistral 3 release) https://x.com/rasbt/statu…

model-releasessebastian-raschka--x

Also kudos to fireworks AI for admitting that they built on top of Kimi (I remember the awkward Mistral 3 release) https://x.com/rasbt/status/1999557830131494977?s=20 Hold on a sec, Mistral 3 Large us

An interesting idea is to continually train specialized models on skills. This paper explores that idea. They propose SkillGym which turns h…

An interesting idea is to continually train specialized models on skills. This paper explores that idea. They propose SkillGym which turns human-written skills into 2,756 training environments with co

model-releasesdair-ai--x

Another 'Harness matters' post (codex cli > pi and opencode)

I run my own LLM while also having a Openai subscription. Also tried DeepSeek (latest flash now). I run Qwen 3.8 flash Next at an amazing speed on my 2x3090 + Ram! But local LLM never did worked for m

b11206

hexagon: support for backend sampler (#29502) hex-topk: trying to improve/cleanup the pipeline hex-sampling: add STEP op hex-sampler: add SUM op hex-sampler: update CPY to support sampling cases hex-b

b11207

hexagon: support tiled Q4_0 and Q8_0 GET_ROWS (#29511) hexagon: support tiled Q4_0 and Q8_0 GET_ROWS hex-get-rows: fix macros hex-get-rows: use tiled HVX dequantization Assisted-by: OpenCode hex-get-r

b11208

hrm : fix layer placement of z_l_init weight (#29512) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/50522371 macOS/iOS: macOS Apple Silicon (arm64) macOS

b11209

llama-bench : fix OOB access of hf_file (#29515) Signed-off-by: Adrien Gallouët angt@huggingface.co Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/50528155

b11211

RPC: use RDMA completion channel to not spin (#29440) RPC: use RDMA completion queue to not spin add TODO for apple RDMA Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/

b11212

common : throw instead of abort on grammar without llguidance (#29516) Signed-off-by: Adrien Gallouët angt@huggingface.co Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp

b11213

vulkan: fix argsort kernel selection for Adreno (#29469) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/50544436 macOS/iOS: macOS Apple Silicon (arm64) mac

b11214

HIP: Enable fattn-mma kernel on cdna for dkq > 256 for large batch sizes (#28907) HIP: Enable fattn-mma kernel on cdna for dkq > 256 for large batch sizes CI: hip-quality-check: ignore spills for very

b11215

CUDA: tune fp16 tile FlashAttention configs for head sizes 40-112 (#26289) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/50554486 macOS/iOS: macOS Apple S

b11216

sycl: FWHT kernels for block widths above 512 (#29243) The SYCL FWHT covers 64 to 512 via the standard butterfly network, plus 384/640/768/1280 via the Kronecker/Paley construction added separately in

b11217

opencl: refine bin kernel loading condition (#29503) Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/50560500 macOS/iOS: macOS Apple Silicon (arm64) macOS A

b11218

jinja : add support for dict builtin (#29477) add support for dict builtin add tests Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestations/50563209 macOS/iOS: mac

b11221

common : make string_split throw on invalid values (#29518) Signed-off-by: Adrien Gallouët angt@huggingface.co Website: https://llama.app Attestations: https://github.com/ggml-org/llama.cpp/attestatio

b11222

common : avoid side effects around params parsing (#29537) register --rpc unconditionally and call llama_supports_rpc() only from its handler print server 'initialization ...' log after args are parse

b11223

server : allow RANK pooling batch splitting for causal LLM rerankers (ie. Qwen3 and Qwen3-VL) (#28876) server : allow splitting RANK pooling for causal LLM rerankers Rerank models fall into two catego