Search: “models”

PetQA: Benchmarking Veterinary Knowledge and Clinical Reasoning

arXiv:2609.04598v1 Announce Type: cross Abstract: We introduce PetQA, a Korean long-form question-answering (QA) benchmark for evaluating veterinary knowledge and clinical reasoning in large language

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Quality Recovery for Quantized KV Caches via Low-Rank Attention Adaptation

DGX agent

arXiv:2609.04263v1 Announce Type: cross Abstract: Low-bit key--value (KV) caches reduce the memory required for autoregressive decoding, but the resulting quality loss depends on the model and quantiz

model-releasesarxiv-cs-cl

7 Sept 2026

Recurrence Is Not Enough: Causally Validating Multilingual SAE Translation Features in Gemma 2 and 3

arXiv:2609.04808v1 Announce Type: cross Abstract: Sparse autoencoder (SAE) features are increasingly used to explain and steer language-model behavior, but it remains unclear whether a feature found i

SciDocBench: A Workflow-Centered Benchmark and Data Pipeline for Scientific Document Understanding

arXiv:2609.05141v1 Announce Type: new Abstract: Scientific papers require models to reason jointly over text, equations, figures, tables, code, and datasets while preserving the provenance of supporti

Single-Query Black-Box Calibration Auditing via Logit Bias

model-releasesarxiv-cs-lg

arXiv:2609.05125v1 Announce Type: new Abstract: Evaluating the calibration of Large Language Models (LLMs) is critical for their safe deployment as zero-shot classifiers. Yet, commercial API providers

Test-Time Adaptation via Cache Personalization for Facial Expression Recognition in Videos

arXiv:2603.21309v4 Announce Type: replace Abstract: Facial expression recognition (FER) in videos requires model personalization to capture considerable variation across subjects. Vision-language mode

TIER: Threat Implicitness Benchmark for Evaluating LLM Safety Behaviors

arXiv:2609.05117v1 Announce Type: cross Abstract: Current LLM safety benchmarks largely rely on binary metrics, overlooking how models respond to harmful prompts with varying threat implicitness. We i

VDN-H3 on M5 Pro 24GB: up to ~2.6× faster H3 generation

model-releasesr-stablediffusion

I recently ported VDN-H3 (Video DeltaNet for MiniMax H3) to Vpipe and tested it on an M5 Pro 24GB. The benefit scales with both sequence length and resolution: at 832×480, VDN-H3 starts to pull ahead

Weather-Conditioned Depth Anything

arXiv:2609.04827v1 Announce Type: new Abstract: Monocular depth estimation foundation models, such as the Depth Anything series, have achieved remarkable performance across diverse domains. However, t

You Really Didn't Get That? Benchmarking Social Pragmatic Inference for Indirect and Playful Chinese Online Comments

arXiv:2609.04384v1 Announce Type: cross Abstract: Chinese online comments often convey social meaning through indirect and playful language that is hard to interpret without context. Existing evaluati

DeepSeek-V4-Flash-Vision Q8 vs Qwen3.8-Flash-Next Q8

I'm using DS-V4-Flash-Vision with Q8_K_XL quantization locally as my everyday engine, and for some time now I've been doing a lot of comparisons with Qwen3.8-Flash-Next, also with Q8_K_XL quantization

6 Sept 2026

Excellent analysis, and yes this happens every single time. (Tyler Cowen said o3 was AGI!)

model-releasesgary-marcus--x

Excellent analysis, and yes this happens every single time. (Tyler Cowen said o3 was AGI!) Every time a new model drops, GPT, Claude, Gemini, Grok, DeepSeek, somebody on this app declares that we have

6 Sept 2026

I released this today because I was 100% sure a frontier AI would figure it out Matter is chiral => the first derivation of the standard mod…

model-releasesemad-mostaque--x

I released this today because I was 100% sure a frontier AI would figure it out Matter is chiral => the first derivation of the standard model & three generations uniquely That’s all, nothing else Can

6 Sept 2026

I've found myself using Local LLM's like 3D printers.

Anyone who has a 3D printer and get use of it finds it incredibly useful for those odd jobs around the house, a missing bracket, a cable router, steam deck holder and so on. In the past if I was missi

5 Sept 2026

Otaku — an LLM frontend

Otaku is an LLM frontend, primarily designed for roleplay, an alternative to SillyTavern and the like. However, It also works for general-purpose chat with local backends (including Ollama) or cloud m

5 Sept 2026

Qwen3.8 27b for agentic coding and next .... what?

First, I'd like to thank the Qwen and Unsloth teams for the Qwen3.8 27b UD Q4_K_XL. Fits the poor 24GB of 3090 VRAM with 100k context at Q8 and works phenomenally well! Imho if theres anything that ca

5 Sept 2026

Research

Accelerating Masked Image Generation by Learning Controlled Latent Dynamics

arXiv:2602.23996v2 Announce Type: replace Abstract: Masked Image Generation Models (MIGMs) have achieved great success, yet their efficiency is hampered by the multiple steps of bi-directional attenti

researcharxiv-cs-cv

Anthropic’s $2 trillion IPO puts powerful external trustees in spotlight

model-releasesars-technica

Anthropic’s planned initial public offering, potentially valuing the company at up to $2 trillion, will be examined under a novel governance model that places an external Long‑Term Benefit Trust (LTBT

Research

Causal-Counterfactual RAG: The Integration of Causal-Counterfactual Reasoning into RAG

arXiv:2509.14435v3 Announce Type: replace Abstract: Large language models (LLMs) have transformed natural language processing (NLP), enabling diverse applications by integrating large-scale pre-traine

researcharxiv-cs-cl

Free Pause Tokens

arXiv:2609.03807v1 Announce Type: cross Abstract: A free pause token gives a language model extra compute to form each next-token prediction (as a pause, or thinking, token does) but carries that comp

I benchmarked 21 Qwen3.8 27B variants on 16GB VRAM

After Qwen3.8 27B came out, I decided to benchmark the models that could fit in my GPU (RTX 5080) on my actual code (C code), the results were not completely unexpected but some quants were definitely

Jina-OCR-v1: Efficient Document Parsing with Speculative Decoding and Dense Verifiable Rewards

arXiv:2609.03181v1 Announce Type: new Abstract: We present Jina-OCR-v1, an end-to-end document parsing model built to serve on low-budget GPUs. It combines the compressed-vision encoder and the 3B mix

Learning What Not to Forget: Long-Horizon Agent Memory from a Few Kilobytes of Learning

arXiv:2606.20954v2 Announce Type: replace-cross Abstract: Long-running language-model systems accumulate interaction history that outgrows the context window, so they must continually evict. When an e

Lost in Reordering: Structural Sensitivity of Multilingual LLMs under Semantics-Preserving Perturbations

arXiv:2609.03511v1 Announce Type: new Abstract: Large Language Models (LLMs) demonstrate strong multilingual reasoning performance, yet their robustness to semantics-preserving structural variation re

MedQA-MM: Shortcuts Behind Medical Visual Reasoning

arXiv:2609.03261v1 Announce Type: cross Abstract: A benchmark score credits final answers, but not the route by which an item can be answered. In medical multimodal multiple-choice questions (MCQs), t

ObserverBench: Testing Mechanistic Estimates for Intervention and Control

arXiv:2609.03026v1 Announce Type: cross Abstract: Mechanistic interpretability is increasingly used to guide interventions such as activation steering, circuit removal, and safety monitoring. Yet an i

Occupancy-based Quantile Risk Control

model-releasesarxiv-cs-lg

arXiv:2609.03104v1 Announce Type: cross Abstract: Conformal risk control is an emerging framework for the safe deployment of machine learning models with finite-sample guarantees. To accommodate a bro

OpenAI's rogue agents were caught communicating via public wikis

model-releasessimon-willison

Here we go again... Discovery of a new OpenAI agent message board by Sydney Von Arx, Cormac Slade Byrd, Spencer Kitts, and Thomas Larsen describes the latest accidental cyberattack by models being tra

Plan Pointers and Record-Directive Form in Budgeted Verification of Inherited Agent Memory

arXiv:2609.03450v1 Announce Type: cross Abstract: An agent that inherits six one-line memories may pull at most one archived source record before acting; a directive written into the store can steer t

Privacy-Preserving Heterogeneous Multi-LLM Federated Inference for Cognitive Diagnosis

arXiv:2609.02947v1 Announce Type: cross Abstract: Significant challenges remain in AI-driven educational systems in balancing privacy preservation with accurate cognitive diagnosis. To overcome this,

Research

Random Attention: Rethinking KV Cache Eviction for Efficient Reasoning

arXiv:2609.03430v1 Announce Type: new Abstract: Large language models achieve superior performance on tasks that require extended reasoning, but long chains of thought make the KV cache a severe memor

researcharxiv-cs-cl

Sensing Which Modality Matters: Evidence-Gated Regularization for Robust VLA Policies

arXiv:2609.03142v1 Announce Type: cross Abstract: Vision-Language-Action (VLA) policies fuse multimodal sensory inputs, but training on limited and homogeneous robot demonstrations encourages spurious

Sharpening the Ensemble: An SSIM-Aligned Residual Refiner for Brain-MRI Inpainting Post-Processing

arXiv:2609.03981v1 Announce Type: new Abstract: Brain-MRI inpainting replaces a masked region of a scan with synthesized, anatomically plausible healthy tissue, so that analysis tools built for health

SPARK: Input-Conditioned Sparse Activation Modulation for Frozen DiT-based Super-Resolution

arXiv:2609.03813v1 Announce Type: new Abstract: Real-world image super-resolution (SR) increasingly relies on Diffusion Transformer (DiT) backbones, whose internal activations can be dominated by a sm

STAIR (STructure Aware Information Retriever): A novel dataset and LLM based retriever for document structure augmentation

arXiv:2609.03874v1 Announce Type: new Abstract: Retrieval Augmented Generation (RAG) is a key component for generating accurate and hallucination free answers using Large Language Models (LLMs). LLMs

SVG-Score: Human-Aligned Evaluation of Text-to-SVG Generation

arXiv:2609.03806v1 Announce Type: new Abstract: Scalable Vector Graphics (SVG) generation is attracting increasing attention as generative models improve in expressiveness and controllability. Progres

TabScope: Question-Adaptive Scope Selection for Table Question Answering

arXiv:2609.03395v1 Announce Type: cross Abstract: Large Language Models (LLMs) have shown strong performance on table question answering, yet their accuracy often degrades as table size increases. We

The Impact of Synthetic Data Augmentation on Discourse-Pragmatic Function Classification

arXiv:2609.03652v1 Announce Type: new Abstract: Synthetic data augmentation has become a common strategy for addressing class imbalance in NLP, but most approaches focus on the quantity and diversity

The Pelican comparison grid for Astra is pretty interesting

model-releasessimon-willison

I got access to GPT-6 Astra this afternoon, so naturally I used it to generate SVGs of pelicans riding bicycles - at low, medium, high, xhigh and max reasoning levels (Astra doesn't support reasoning=

ThousandWorlds: A benchmark for climate emulation of potentially habitable exoplanets

model-releasesarxiv-cs-lg

arXiv:2606.18338v2 Announce Type: replace Abstract: The search for life beyond Earth will depend on detecting faint signatures in the atmospheres of potentially habitable exoplanets. Interpreting thos

Toward Collective-Centric Evaluation of Preference Inference for Participatory Democracy

arXiv:2609.02990v1 Announce Type: cross Abstract: To scale up collective decision-making, participatory democracy platforms such as Polis and Remesh enable online deliberation among thousands of parti

Local Ai

You can now run a 90M conversational LLM on the Sony PSP (hardware from 2004). Doesn't get more local than this.

Github link: https://github.com/thatblend/LLMPSP I wanted to see what the PSP can theoretically handle and I got my answer - a 90M model is about the max it can do without atrocious inference speeds.

local-air-localllama

APEx: Distillation of Agent Procedural Experience for Adaptive Deep Research Question Answering

arXiv:2609.02253v1 Announce Type: new Abstract: Deep research agents augment large language models with external tools to answer complex, long-horizon questions through multi-turn reasoning. Learning

b10776

model-releasesllama-cpp-releases

model: add NVIDIA Nemotron-3-Puzzle-75B-A9B (NemotronHPuzzle) support (#25444) hparams: add per-layer n_ff_exp/n_expert_used arrays with scalar-or-array loading G1/G2 infrastructure for variable-per-l

Benchmarking RAW and RGB Restoration in Image Signal Processors

arXiv:2609.02831v1 Announce Type: new Abstract: Modern cameras transform RAW sensor measurements into sRGB images through an image signal processor (ISP). We benchmark two placements for blind restora

Bridging Latent Reasoning and Target-Language Generation via Retrieval-Transition Heads

arXiv:2602.22453v3 Announce Type: replace Abstract: Recent work has identified a subset of attention heads in Transformer as retrieval heads, which are responsible for retrieving information from the

C^{3}T: Counterfactual Causal Reasoning for Sentiment Shifts in Social-Media Conversation Trees

arXiv:2609.02131v1 Announce Type: cross Abstract: Sentiment in social-media threads does not only vary across posts; it shifts as users react to claims, corrections, evidence, and hostility within a b

CORAL: An LLM-Native Harness for Production Recommender Systems

arXiv:2609.02730v1 Announce Type: new Abstract: Production recommender systems shape what billions of people see, and sustaining their performance requires continual optimization: as content, user beh