AllLLMRoboticsStartupsResearchEthicsBusinessHardwareAutonomousMedicineCreative
Topic

#LLM

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

4835 articles-77% vs prev. week320 this week
Subscribe
RELATED:#Business#Hardware#Research#Creative#Startups#Ethics#Autonomous#Robotics#Medicine
ForcingBooster Halves Denoising Steps in Autoregressive Video Generation
LLM·arXiv cs.CV

ForcingBooster Halves Denoising Steps in Autoregressive Video Generation

Oct 12, 2026·2 min
Video-Language Models Confuse Car Maneuvers: LoRA and Probes Fix the Error
LLM

Video-Language Models Confuse Car Maneuvers: LoRA and Probes Fix the Error

Oct 12, 2026·2 min
TerraAgent checks whether the AI actually looks at the image: hallucinations drop from 17.1% to 9.8%
LLM

TerraAgent checks whether the AI actually looks at the image: hallucinations drop from 17.1% to 9.8%

Oct 12, 2026·2 min
AVA improves multimodal ICL accuracy by 1.7 points without any training
LLM

AVA improves multimodal ICL accuracy by 1.7 points without any training

Oct 12, 2026·2 min
Coral-KV Compresses Video Model KV-Cache by 37–48% and Lifts PSNR by Up to 4.95 dB
LLM·arXiv cs.CV

Coral-KV Compresses Video Model KV-Cache by 37–48% and Lifts PSNR by Up to 4.95 dB

Researchers propose Coral-KV, a KV-cache compression method for autoregressive video generation. On MAGI-1 it gains 4.95 dB PSNR over QVG-4 while using 37% less…

Oct 12, 2026·2 min
Black Lake: Chinese Factories Skip Digitization and Go Straight to AI Agents
LLM·SCMP Tech

Black Lake: Chinese Factories Skip Digitization and Go Straight to AI Agents

Black Lake CEO Zhou Yuxiang likens factories' shift to AI agents to China's leap from bank cards to mobile payments, and plans to export the approach to global…

Oct 12, 2026·2 min
T. Rowe Price: 55% of Hong Kong Residents Expect AI to Hit Their Retirement, but Only One in Five Feels Ready
LLM·SCMP Tech

T. Rowe Price: 55% of Hong Kong Residents Expect AI to Hit Their Retirement, but Only One in Five Feels Ready

A T. Rowe Price survey of 1004 Hong Kong residents: 55% expect AI to destabilize their careers and income, but only one in five is financially prepared for that…

Oct 12, 2026·2 min
Genetic Optimizer Cuts Aerosol Sensor Placement Time in the UK by 60%
LLM·arXiv cs.LG

Genetic Optimizer Cuts Aerosol Sensor Placement Time in the UK by 60%

Researchers built a neural process into DeepSensor and cut the time to compute aerosol monitoring station placement in the UK from 25 to 10 minutes.

Oct 12, 2026·2 min
Top themes
01Kronecker Product Attention Compresses the KV-Cache Up to 10x and Matches MLA on QualityOct 12, 2026·2 min02DiffusionART speeds up RL fine-tuning of video models 12x, with no EMA tuningOct 12, 2026·2 min03A Physical Transformer Built Entirely From Transistors Learns In-Context Learning in SimulationOct 12, 2026·2 min04Theory Shows Masked Diffusion on Transformers Learns from Õ(LK/ε) SamplesOct 12, 2026·2 min05SAZO speeds up LLM fine-tuning without backprop: up to 6× fewer queries to reach the same lossOct 12, 2026·2 min
Holomorphic Neural Networks With Singular Functions Beat All Methods on Laplace Benchmarks
LLM·arXiv cs.LG

Holomorphic Neural Networks With Singular Functions Beat All Methods on Laplace Benchmarks

Oct 12, 2026·2 min
CacheScope speeds up diffusion models by up to 3.57x and cuts LPIPS by 70%
LLM·arXiv cs.LG

CacheScope speeds up diffusion models by up to 3.57x and cuts LPIPS by 70%

Oct 12, 2026·2 min
CircuitTree assembles 39 HuBERT phoneme circuits into a syllable transcriber with 64.8% accuracy
LLM·arXiv cs.LG

CircuitTree assembles 39 HuBERT phoneme circuits into a syllable transcriber with 64.8% accuracy

Oct 12, 2026·2 min
ResonatorLM Taught to Read Memory by Token: Perplexity at 1024 Drops from 12.81 to 3.76
LLM·arXiv cs.LG

ResonatorLM Taught to Read Memory by Token: Perplexity at 1024 Drops from 12.81 to 3.76

Oct 12, 2026·2 min
Kairos: a byte-level model for all modalities converges on 440 bytes
LLM·arXiv cs.LG

Kairos: a byte-level model for all modalities converges on 440 bytes

Kairos' author built a byte-level multimodal architecture with MoE and linear attention and checked that it trains at all: loss of about 10⁻³ on a single 440-byte…

Oct 12, 2026·2 min
Open LLMs Explain Satire More Cheaply Than Frontier Models, but Handle Australian Satire Worse Than American
LLM·arXiv cs.CL

Open LLMs Explain Satire More Cheaply Than Frontier Models, but Handle Australian Satire Worse Than American

An arXiv pilot study: an "unreliable narrator" framing makes open models a workable, cheaper alternative to frontier systems for explaining satire. Australian texts…

Oct 12, 2026·2 min
COFFEQA Adds 8.77 EM on TriviaQA to Few-Shot Distillation Using Counterfactuals
LLM·arXiv cs.CL

COFFEQA Adds 8.77 EM on TriviaQA to Few-Shot Distillation Using Counterfactuals

COFFEQA pairs hard examples with counterfactuals. On TriviaQA with 32 pairs, it beats baseline distillation for T5 by 8.77 EM and 9.98 F1.

Oct 12, 2026·2 min
MADS Trains Dialog Agents Not to Break: 1,095 Conversations, 66% With "Bad" Scenarios
LLM·arXiv cs.CL

MADS Trains Dialog Agents Not to Break: 1,095 Conversations, 66% With "Bad" Scenarios

Oct 12, 2026·2 min
U4RASD Takes 1st Place at KnowledgeGraphEval-2026: 94.06 micro-F1 on Arabic
LLM·arXiv cs.CL

U4RASD Takes 1st Place at KnowledgeGraphEval-2026: 94.06 micro-F1 on Arabic

Oct 12, 2026·2 min
AXLE Method Primes Annotators' Own Experience Before Hate Speech Labeling: 104 People, 1,040 Ratings
LLM·arXiv cs.CL

AXLE Method Primes Annotators' Own Experience Before Hate Speech Labeling: 104 People, 1,040 Ratings

Oct 12, 2026·2 min
perturber: a service with 68 text perturbations tests LLM robustness
LLM·arXiv cs.CL

perturber: a service with 68 text perturbations tests LLM robustness

Oct 12, 2026·2 min
GeopoliticalQA: bge-m3 Beats Commercial Embeddings on 5 525 Report-Based Questions
LLM·arXiv cs.CL

GeopoliticalQA: bge-m3 Beats Commercial Embeddings on 5 525 Report-Based Questions

Oct 12, 2026·2 min
Qwen Picks What to Remember: Replay From a Frozen Model Reduces Forgetting
LLM

Qwen Picks What to Remember: Replay From a Frozen Model Reduces Forgetting

Oct 12, 2026·3 min
CLUE Builds an Event Ontology Without Experts in Three Steps: Embeddings, Clusters, Names
LLM

CLUE Builds an Event Ontology Without Experts in Three Steps: Embeddings, Clusters, Names

Oct 12, 2026·2 min
AutoIndexer Teaches a Language Model to Edit Its Own Output with Three Types of Edits
LLM

AutoIndexer Teaches a Language Model to Edit Its Own Output with Three Types of Edits

Oct 12, 2026·2 min
Seq-Drifting Generates Text in a Single Pass: Perplexity of 36.01 on LM1B
LLM·arXiv cs.AI

Seq-Drifting Generates Text in a Single Pass: Perplexity of 36.01 on LM1B

Researchers propose Seq-Drifting: a model turns Gaussian noise into text in a single forward pass and reaches a generative perplexity of 36.01 on the LM1B dataset.

Oct 12, 2026·2 min
CToM teaches LLMs to predict user replies and lifts accuracy to 79.6%
LLM·arXiv cs.AI

CToM teaches LLMs to predict user replies and lifts accuracy to 79.6%

The new CToM method has a model write a draft, predict the user's reaction, and rewrite its answer. Math accuracy rose from 74.1% to 79.6%.

Oct 12, 2026·2 min
Related topics
Business 1158Research 497Robotics 178Hardware 101Autonomous 17Startups 9Medicine 8Ethics 5