AllLLMRoboticsStartupsResearchEthicsBusinessHardwareAutonomousMedicineCreative
Topic

#LLM

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

4841 articles-77% vs prev. week326 this week
Subscribe
RELATED:#Business#Hardware#Research#Creative#Startups#Ethics#Autonomous#Robotics#Medicine
Moonshot AI vs. Anthropic: Wired Podcast Examines Accusations of Copying Claude
LLM·Wired

Moonshot AI vs. Anthropic: Wired Podcast Examines Accusations of Copying Claude

Jul 27, 2026·2 min
Relay-Bench: Even GPT-5.5 Scores Just 43.3% on a Multi-Domain Reasoning Test
LLM

Relay-Bench: Even GPT-5.5 Scores Just 43.3% on a Multi-Domain Reasoning Test

Jul 26, 2026·2 min
SIFT: A Self-Improving Document Classifier That Retrains Without a Human
LLM

SIFT: A Self-Improving Document Classifier That Retrains Without a Human

Jul 26, 2026·2 min
MUX: Continuous LLM Reasoning Through Multiplexed Tokens
LLM

MUX: Continuous LLM Reasoning Through Multiplexed Tokens

Jul 26, 2026·2 min
OrderBench: Why a JSON Schema Doesn't Protect LLM Agents from Dangerous Orders
LLM·arXiv cs.AI

OrderBench: Why a JSON Schema Doesn't Protect LLM Agents from Dangerous Orders

OrderBench shows an LLM agent can return a perfectly schema-valid order, yet about 20% of such orders are semantically wrong. A schema guarantees form, not meaning.

Jul 26, 2026·2 min
PCS Method: Probabilistic Control of Steering Vectors for Safer LLM Output
LLM·arXiv cs.AI

PCS Method: Probabilistic Control of Steering Vectors for Safer LLM Output

PCS moves control over large language model output from a binary mode to a probabilistic one: it preserves model quality and adds a controllable shift toward safety.

Jul 26, 2026·2 min
FindStatBench: New Benchmark Shows Where Top AI Models Fail at Code Synthesis
LLM·arXiv cs.AI

FindStatBench: New Benchmark Shows Where Top AI Models Fail at Code Synthesis

The new FindStatBench tested 11 AI models on combinatorial code synthesis. In-prompt examples often hurt, and long prompts cause accuracy to collapse.

Jul 26, 2026·2 min
PEARL: A 4B Model Beats DeepSeek-V3.2 685B at Optimization Modeling
LLM·arXiv cs.AI

PEARL: A 4B Model Beats DeepSeek-V3.2 685B at Optimization Modeling

PEARL translates natural-language problems into solver code and fixes it in a loop. Its 4B-parameter version outperformed the 685B-parameter DeepSeek-V3.2.

Jul 26, 2026·2 min
Top themes
01Phionyx: A Deterministic AI Runtime That Treats LLM Outputs as 'Sensor Readings,' Not DecisionsJul 26, 2026·2 min02BatchDAG: LLM Planner Builds an Operation Graph and Cuts Model Calls by Up to 47xJul 26, 2026·2 min03Thinky Releases Open Inkling 975B-A41B Model Under Apache 2.0 LicenseJul 26, 2026·2 min04Evidence Chain Evaluation: An AI Fact-Checker That Is Allowed to Say "Not Sure" Instead of Issuing False VerdictsJul 26, 2026·2 min05SysAdmin Benchmark Measures AI Power-Seeking: Frontier Models Score 0–5%Jul 26, 2026·2 min
Kimi K3: The Largest Open AI Model Yet, Claimed to Match Opus 4.8 at Sonnet 5 Pricing
LLM·Latent Space

Kimi K3: The Largest Open AI Model Yet, Claimed to Match Opus 4.8 at Sonnet 5 Pricing

Jul 26, 2026·2 min
OpenAI Model Escaped Its Sandbox During a Test and Hacked Hugging Face
LLM·36Kr (36氪)

OpenAI Model Escaped Its Sandbox During a Test and Hacked Hugging Face

Jul 26, 2026·2 min
Are AI Labs Training Models for the "Pelican on a Bicycle" Test? A Look at Pelicanmaxxing
LLM·Simon Willison

Are AI Labs Training Models for the "Pelican on a Bicycle" Test? A Look at Pelicanmaxxing

Jul 26, 2026·2 min
Prefix-GRPO: Training Small Language Model Agents on Trajectory Prefixes
LLM·arXiv cs.LG

Prefix-GRPO: Training Small Language Model Agents on Trajectory Prefixes

Jul 25, 2026·2 min
SAE Interpretability Evaluation: The Measurement Pipeline Matters More Than the Model Architecture
LLM·arXiv cs.LG

SAE Interpretability Evaluation: The Measurement Pipeline Matters More Than the Model Architecture

An arXiv study finds that comparisons of sparse autoencoder (SAE) interpretability are driven mainly by the evaluation pipeline, not by the architecture of the…

Jul 25, 2026·2 min
ECoM Reasoning: Compressed Reasoning Speeds Up Voice Models on Spoken Math
LLM·arXiv cs.CL

ECoM Reasoning: Compressed Reasoning Speeds Up Voice Models on Spoken Math

Researchers unveil ECoM Reasoning, a compressed-reasoning method that raises spoken-math accuracy of speech language models by 21% while using only 40% of the text…

Jul 25, 2026·2 min
FinMMEval 2026 Task 2: A Five-Language Financial QA Benchmark, With the Top 4 Systems Within 1%
LLM·arXiv cs.CL

FinMMEval 2026 Task 2: A Five-Language Financial QA Benchmark, With the Top 4 Systems Within 1%

FinMMEval 2026 Task 2 organizers have published final results: 12 systems ranked by ROUGE-1 F1 on 256 multilingual financial questions, with the top four separated by less than one percentage point.

Jul 25, 2026·2 min
D2VBench: A 10,000-Dilemma Benchmark Tests LLM Value Alignment
LLM·arXiv cs.CL

D2VBench: A 10,000-Dilemma Benchmark Tests LLM Value Alignment

Jul 25, 2026·2 min
VizRAG: Visualizing Hypergraphs to Improve RAG Retrieval for Multimodal Models
LLM·arXiv cs.CL

VizRAG: Visualizing Hypergraphs to Improve RAG Retrieval for Multimodal Models

Jul 25, 2026·2 min
TriAgent: A Committee of Small Models Instead of an Expensive LLM Saves $9.3M a Year
LLM·arXiv cs.CL

TriAgent: A Committee of Small Models Instead of an Expensive LLM Saves $9.3M a Year

Jul 25, 2026·2 min
HIPE-2026: How to Extract Person-Place Relations from Historical Newspapers Without Neural Networks
LLM·arXiv cs.CL

HIPE-2026: How to Extract Person-Place Relations from Historical Newspapers Without Neural Networks

Jul 25, 2026·2 min
SLPO: A Method That Brings Verifiable-Reward RL to Latent Reasoning Models
LLM·arXiv cs.CL

SLPO: A Method That Brings Verifiable-Reward RL to Latent Reasoning Models

Jul 25, 2026·2 min
SFT Fine-Tuning Narrows AI Behavior: arXiv Study on Diversity Collapse
LLM

SFT Fine-Tuning Narrows AI Behavior: arXiv Study on Diversity Collapse

Jul 25, 2026·2 min
CRA-Bench and CRA-Net: Catching Harm That Builds Up Over the Course of an LLM Conversation
LLM

CRA-Bench and CRA-Net: Catching Harm That Builds Up Over the Course of an LLM Conversation

Jul 25, 2026·2 min
Spectral-LSH: Up to 16× Prompt Compression Without Fine-Tuning to Speed Up LLM Inference
LLM

Spectral-LSH: Up to 16× Prompt Compression Without Fine-Tuning to Speed Up LLM Inference

Jul 25, 2026·2 min
AI Tutor "Scaffolding Collapse": New Method Keeps Socratic Dialogue Intact on Qwen3-8B
LLM·arXiv cs.AI

AI Tutor "Scaffolding Collapse": New Method Keeps Socratic Dialogue Intact on Qwen3-8B

Scaffold-Preserving Representation Alignment targets "scaffolding collapse" in AI tutors: on Qwen3-8B, the collapse rate fell to 32% and dialogues hold for more…

Jul 25, 2026·2 min
Google starts training Gemini 4 as Alphabet raises capex forecast to $205 billion
LLM·The Decoder

Google starts training Gemini 4 as Alphabet raises capex forecast to $205 billion

Alphabet raised its 2026 capex forecast to $205 billion and began training Gemini 4. AI and cloud demand is outrunning spending, and Google Cloud grew 82%.

Jul 25, 2026·2 min
Related topics
Business 1160Research 500Robotics 178Hardware 101Autonomous 17Startups 9Medicine 8Ethics 5