AI Research
Oct 10, 2026 · 4 min read
OpenAI Retracts Three Math Papers After Sign Error in 722-Manuscript Release
Three mathematics papers collapsed because of one wrong sign. OpenAI retracted them within 24 hours of publishing 722…
AI Agents
Oct 9, 2026 · 3 min read
TGIT Bridges the Gap Between Human Speech and Frozen Aerial Drones
Aerial drones can thread through complex three-dimensional environments with increasing confidence, but ask one to "fly around the…
AI Research
Oct 9, 2026 · 5 min read
Unified Time-Series AI Model Tops 3 Benchmarks With 27B Qwen Backbone
A paper posted to arXiv on October 1, 2026 by researchers from Columbia, Stanford and Google Agentic Systems…
AI Research
Oct 9, 2026 · 4 min read
Google SynthID Detector Verifies AI Media Across OpenAI and Nvidia Models
Google DeepMind's SynthID Detector went public on October 7, 2026, and it works across content from OpenAI and…
AI Research
Oct 5, 2026 · 5 min read
CVaR-GPA Enhances Generative AI for Tail Accuracy
A market crash, a catastrophic flood, a novel pathogen: these are precisely the events that standard generative models…
AI Research
Oct 3, 2026 · 5 min read
AstroReason-Bench Shows LLM Agents Fall Short in Space Mission Planning
Current LLM agents cannot reliably plan a space mission. AstroReason-Bench, a benchmark released in January 2026, puts that…
AI Research
Oct 2, 2026 · 5 min read
ICML 2026 Tightens Rules as LLM Policy Shows Near-Zero Effect on Scores
A randomised study posted to arXiv on September 16, 2026 tested whether ICML 2026’s LLM review policies actually…
AI Research
Oct 1, 2026 · 4 min read
SimpleEvol’s LLM-Agents Automate Heuristic Design with Fewer Priors
Most AI research adds complexity. The SimpleEvol paper, posted to arXiv on September 30, 2026, argues the opposite:…
AI Research
Sep 30, 2026 · 5 min read
LLM Agents: User-Side Model Solves Recommendation Cold-Start and Filter Bubbles
The cold-start problem and filter bubbles have long been treated as engineering challenges to optimise around. A paper…
AI Research
Sep 27, 2026 · 4 min read
River AI Recipe Boosts Terminal Agent Generalization 106%
A training recipe called River has doubled generalisation performance for AI terminal agents, and does it with fewer…