All LLM Robotics Startups Research Ethics Business Hardware Autonomous Medicine Creative

Topic

#LLM

Curated coverage on «LLM»: releases, research, benchmarks, open-source.

4813 articles-74% vs prev. week354 this week

LLM·Jiqizhixin (机器之心)

WorldMemBench Tests Whether World Models Remember Rooms and Events

Oct 23, 2026·2 min

LLM

University of Pittsburgh Looks for a 3D Map Inside a VLM: How a Model Sees a Room

Oct 23, 2026·2 min

LLM

LightSprint Raises $6.1M for a Multi-Agent Product-to-Engineering Loop

Oct 23, 2026·2 min

LLM

Cohere proposes a fix for agent sprawl: competitive selection instead of bans

Oct 23, 2026·2 min

LLM·Hugging Face Daily Papers

ACEdit Teaches Image Editors to Train on Their Own Mistakes Without Degrading

ACEdit applies Self-Forcing to image editing: the model learns from its own outputs, while a teacher sees the original image and a composite instruction.

Oct 23, 2026·2 min

LLM·Hugging Face Daily Papers

Mine-Sample-Render: A Three-Stage Method for Diverse Generation

Modern generative models are high-quality but monotonous. MSR (Mine-Sample-Render) is a Microsoft Research approach for diverse generation at test time. An explicit…

Oct 23, 2026·2 min

LLM·Hugging Face Daily Papers

Sparse Autoencoders: Pairwise Interactions Lift Rank Fidelity to 0.7790

The new SAR metric shows SAE features add up better than a random basis in all 18 settings, but accurate steering still requires pairwise interactions.

Oct 23, 2026·2 min

LLM·Hugging Face Daily Papers

PathFM: generating complete time series trajectories through flow matching

Autogluon researchers have introduced PathFM, a model that generates complete time series trajectories instead of simple point forecasts. On a dedicated benchmark…

Oct 23, 2026·2 min

Top themes

01Weight editing without retraining cut sexist responses from 57.9% to 24.3%Oct 23, 2026·2 min 02Dusting the Phantom: smart Gaussian allocation for dynamic scenesOct 23, 2026·1 min 03DuET cuts piano-to-score transcription error from 6.56% to 3.64%Oct 23, 2026·2 min 04LexFrame reads LLM hidden states: 0.97 coherence against a lens ceiling of 0.30Oct 23, 2026·3 min 05OmniXRef: A Unified Framework for Video Editing with Multimodal ReferencesOct 23, 2026·2 min

LLM·Vercel Blog

Vercel Launches Python in General Availability: Six SDKs and FastAPI Support

Oct 23, 2026·2 min

LLM·Jiqizhixin (机器之心)

Matrix Game 4.0 Hits 60 FPS on a Single Gaming GPU and Beats Nvidia's Rival Model

Oct 23, 2026·2 min

LLM·Jiqizhixin (机器之心)

CodeActionBench: GPT-6 Astra in Codex CLI Hits 73.3% Success on Two-Handed Tasks

Oct 23, 2026·2 min

LLM·YouTube: Fireship

Cloudflare acquires Deno, and the JavaScript runtime war tilts toward Node

Oct 23, 2026·2 min

LLM·Jiqizhixin (机器之心)

Stanford and Together AI Teach LLMs to Write Text Starting From an Outline: Generation Up to 31x Faster

Stanford and Together AI researchers brought VAR's "next-scale prediction" to text: the model writes an outline first, then fills in details. The claimed speedup is…

Oct 23, 2026·2 min

LLM·Jiqizhixin (机器之心)

DeepSeek credits users with tokens as compensation after multi-hour outage

On the morning of October 15, DeepSeek's web chat and API, including V4 Pro and V4.1 Flash, were down for several hours. The company made a rare move for the…

Oct 23, 2026·2 min

LLM·Hugging Face Daily Papers

Scene-Designer Learns to Design and Critique 3D Scenes from Text Using 50,000 Examples

Scene-Designer learns to design and critique 3D scenes from text. Two datasets of 35,000 and 15,000 examples. Scene-Critic hits 89% accuracy in scene comparisons, outperforming GPT-5.2.

Oct 23, 2026·2 min

LLM·Hugging Face Daily Papers

RAP compiles 25,000 paper-poster pairs to teach AI how to lay out posters in HTML/CSS

Oct 23, 2026·1 min

LLM·Jiqizhixin (机器之心)

Meta to Use Muse Spark 1.3 in Medical Research, Promises a 10x Speedup

Oct 23, 2026·2 min

LLM·Jiqizhixin (机器之心)

Apple and CMU Speed Up AI Agents 10x by Replacing Reasoning with Selection

Oct 23, 2026·2 min

LLM·Jiqizhixin (机器之心)

KTransformers Moves Part of the MoE Computation to the CPU and Eliminates Weight Transfers During Fine-Tuning

Oct 23, 2026·2 min

LLM·Jiqizhixin (机器之心)

UniAVC: Joint Audio-Video Synthesis via Asymmetric Noise Scheduling

Researchers at HKUST have developed UniAVC, the first model for simultaneous audio-video synthesis. The system trains on each modality's data separately, then…

Oct 23, 2026·2 min

LLM·Groq Blog

Groq Launches Meta Muse Spark Open: 816B Parameters and 350+ Tokens/s

Meta released Muse Spark Open, an open-weight model for agentic coding: 816B parameters, 42B active. It is available on GroqCloud in limited preview at 350+…

Oct 23, 2026·2 min

#LLM

WorldMemBench Tests Whether World Models Remember Rooms and Events

University of Pittsburgh Looks for a 3D Map Inside a VLM: How a Model Sees a Room

LightSprint Raises $6.1M for a Multi-Agent Product-to-Engineering Loop

Cohere proposes a fix for agent sprawl: competitive selection instead of bans

ACEdit Teaches Image Editors to Train on Their Own Mistakes Without Degrading

Mine-Sample-Render: A Three-Stage Method for Diverse Generation

Sparse Autoencoders: Pairwise Interactions Lift Rank Fidelity to 0.7790

PathFM: generating complete time series trajectories through flow matching

Vercel Launches Python in General Availability: Six SDKs and FastAPI Support

Matrix Game 4.0 Hits 60 FPS on a Single Gaming GPU and Beats Nvidia's Rival Model

CodeActionBench: GPT-6 Astra in Codex CLI Hits 73.3% Success on Two-Handed Tasks

Cloudflare acquires Deno, and the JavaScript runtime war tilts toward Node

Stanford and Together AI Teach LLMs to Write Text Starting From an Outline: Generation Up to 31x Faster

DeepSeek credits users with tokens as compensation after multi-hour outage

Scene-Designer Learns to Design and Critique 3D Scenes from Text Using 50,000 Examples

RAP compiles 25,000 paper-poster pairs to teach AI how to lay out posters in HTML/CSS

Meta to Use Muse Spark 1.3 in Medical Research, Promises a 10x Speedup

Apple and CMU Speed Up AI Agents 10x by Replacing Reasoning with Selection

KTransformers Moves Part of the MoE Computation to the CPU and Eliminates Weight Transfers During Fine-Tuning

Kling 4.0 Launches With Native 30-Second Video Generation and Combat Scenes

Cognition and National Design Studio Launch American Forge, a 6-Month Residency

ProD Stops AI Agents From Looping: Tested on 8 Models (0.6B–8B)

ECHO: Multi-Step Forecasting Without Add-Ons, Half the Cost of MTP

UniAVC: Joint Audio-Video Synthesis via Asymmetric Noise Scheduling

Groq Launches Meta Muse Spark Open: 816B Parameters and 350+ Tokens/s