Zhemal Khamidun

Research

PoS Method Catches AI Agents Stuck in Loops of False Belief: 36 Repeats in ALFWorld

An ALFWorld agent forgot to heat a cup and checked the cabinet 36 times in a row until its action budget ran out. A new paper calls this Belief Trapping and…

Robotics·Jiqizhixin (机器之心)

Ego Data Gets a "Full Body": A New Scaling Path for World Models

Robot data collection is shifting from expensive platforms to lightweight formats. Jiqizhixin examines the idea of extending ego recordings with whole-body motion…

Business·Google DeepMind Blog

Google DeepMind and Singapore Release Gemma-Based Scam Detector as Fraud Losses Top S$4 Billion

GovTech and Google DeepMind fine-tuned Gemma on local scam patterns. The model runs on a smartphone, reads chats without the cloud, and warns users before they send…

SAE-IDF Matches Symbols Across Different Systems and Beats a Baseline by Up to 45 Points

SAE-IDF uses sparse autoencoders to match elements across heterogeneous symbolic systems. It beats a dense baseline on 12 tasks with Gemma-2-9B, winning by up to 45…

01ReaDy speeds up hybrid LLM decoding 3.16x on a single B200Oct 14, 2026·2 min 02DAGFormer Plans Connections Between Layers and Cuts Perplexity by 17–25%Oct 14, 2026·2 min 03Researchers Find a 0.4–1.1% Window in Pythia Pretraining Where the Readout ReorganizesOct 14, 2026·2 min 04semCER Weights OCR Errors by Severity: 61M LLM-Judge RatingsOct 14, 2026·2 min 05SAE on Gemma-2-2B: Share of Matched Features Rises from 2% to 29% With 50M TokensOct 14, 2026·2 min

Top themes

REASON teaches a model to think in two loops and saves about half the compute

Robotics·Artificial Analysis

AA-Robotics v0.1: GPT-6 Astra controls a robot arm and scores 69 out of 100

LLM·Hugging Face Daily Papers

MULTIFINGROUND: LLMs Recognize Accounting Concepts With 99% Accuracy but Miss Facts

LLM·Hugging Face Daily Papers

Interlock teaches agents to finish the job: up to +6.64 points on JobBench

BlenderWorld: GPT-6-Astra Beats Opus-5 in 3D Benchmark with a Score of 79.9

The new BlenderWorld benchmark ran 13 frontier models through 223 3D creation tasks in Blender. GPT-6-Astra leads with 79.9, but spatial coherence still lags.

Business·Modal Blog

Modal raises $750M at $15.75B valuation and reaches $1B run rate

The AI infrastructure startup tripled usage in five months, passed $1 billion in annualized run rate and closed a $750 million Series D led by Accel.

LLM·Cohere Blog

Cohere finds coding-agent teams can get 61.8% of the reward for $155 instead of $510

Cohere tested how to organize a team of coding agents on ProgramBench. A strong lead with cheap workers delivers near-frontier results at 70% lower cost.

Robotics·Hugging Face Daily Papers

WinARDet fixes looping in MLLM detectors: F1 rises from 32.16 to 39.36

CyberArena Pits 12 AI Agents Against Each Other in 466 Real-Time Attack-and-Defense Matches

LLM·YouTube: Google DeepMind

Google Uses Gemini 4 Argon to Refactor Chromium and Its Millions of Lines of Code

LLM·YouTube: Google DeepMind

Gemini 4 Argon Found Over Twice as Many Vulnerabilities as Baseline Models in Wiz Tests

Oct 13, 2026·3 min

Robotics·YouTube: Google DeepMind

Gemini 4 Argon Teaches Apollo 2 Humanoid to Play Cornhole: 7 Hits in a Row

Business

Isomorphic Labs Uses Gemini 4 Argon to Parse 500-Page Patents in a Day Instead of a Month

LLM

Habr Author Tests How Many Edits It Takes to Erase an LLM Watermark

Hardware

Wits Physicists Build a "Light Brain" That Reads Hidden Patterns in Light

Business·Qdrant Blog

Perplexity releases two late interaction models for searching PDF pages without OCR

Perplexity has open-sourced pplx-embed-v2-late: two late interaction models that index a PDF page as an image, tables, charts and layout included, with no OCR step.

Business·GitHub Changelog

GitHub adds AI Scan metrics for pull requests to security overview

GitHub organization and enterprise admins can now see in security overview how many findings AI Scan raised in pull requests and how many were resolved.