OpenAI mistranslated mathematics into code for its Navier-Stokes proof
New Scientist says OpenAI mistranslated math into code for its Navier-Stokes proof.
Stories per day · 30 days
43 RSS
403 stories, newest first · one card per story
Stories tagged #ai-research
"As a Language Model": Chat Template Switches LLM Self-Referential Voice
An arXiv paper finds chat templates change how LLMs use self-referential voice.
Beyond Compression: Training Latent Representations for Stable Long-Horizon Rollout in Neural Surrogate Solvers
Research on improving long-horizon rollout stability in neural surrogate solvers, reducing error by 40% while matching full-resolution models on physics benchmarks.
AERIAL: Adversarial Evaluation of Robustness in Accuracy-Preserving Low-Precision EEG Decoders
A research paper on low-precision EEG decoders examines adversarial robustness across compression techniques.
Inside Basecamp Research, the AI startup turning evolution into training data
AI startup Basecamp Research raised $140M from Nvidia and Anthropic to train biological AI models on massive novel genetic datasets for drug discovery.
3
Trust in Edge-Enabled IoT Security: Features, Challenges and Research Directions
Reviews current trust management in edge-enabled IoT, proposes taxonomy, and identifies gaps for future research.
CARE: Experience-Guided Atomic Corrective Execution for Vision-Language-Action Policies
Vision-Language-Action (VLA) policies achieve strong performance in robotic manipulation but remain brittle once execution deviates from nominal trajectories. We propose CARE (Corrective Atomic Robotic Execution), a…
1
Why don't machine learning research agents overfit?
Amazon researchers explain why ML research agents avoid benchmark overfitting, attributing generalization to compressibility of successful strategies.
Why you should work on AI for AI Research — Richard Socher of Recursive
Richard Socher's new lab Recursive, backed by $4.65B seed, targets AI systems that automate AI research itself.
3
SiliconBench: Speed, Memory, and Fidelity for LLM Serving on Unified-Memory Desktops
Schrödinger's Code Repository: Have LLMs Learned SWE-bench or Memorized It?
Lessons from the hacks
11 min readai-research
1
Training-Free Speech-Centric Omni Understanding with Frozen VLMs
AI cyber capability is speeding past earlier projections
2 min readai-research
8
AWS to retire Shield Advanced L7 automatic mitigation on January 1, 2027
4 min readai-research
1
Nearly half of open-source AI projects never reach production
3 min readai-research
2
NVIDIA goes open source with a big batch of physical AI agent tools
3 min readai-research
2
macOS is becoming a proving ground for AI agents
5 min readai-research
4Crusoe brings serverless fine-tuning to AI model development
4 min readai-research
2
The uptime questions every engineering leader should ask this week
9 min readai-research
2
Sycophantic chatbots and the harms that build over many chats
4 min readai-research
2
AWS, Wasabi, Cloudflare, and Backblaze go head-to-head in new cloud storage test
4 min readai-research
2
Intel puts its data center performance knowledge on GitHub
2 min readai-research
3
GitHub releases an open dataset for multilingual developer content
2 min readai-research
3