Research

Princeton study finds Claude and GPT fail at AI research

The Decoder • 4 days ago

Axiom Math AI Verifies Complex 246 Theorem Proof

IEEE Spectrum AI • 1 day ago

World Labs Unveils R2S2R Simulation Engine for Robots

The Decoder • 3 days ago

Suppressing Self-Reflection Shifts Llama and Gemma Beliefs

The Decoder • 2 days ago

Text Color Changes Can Manipulate Qwen2-VL and Other VLMs

Unite.AI • 1 day ago

KIVI Quantization Secures Llama-2 Accuracy at Two Bits

Unite.AI • 4 days ago

NVIDIA Shrinks Nemotron 3.5 Lightning to 22 GB

NVIDIA Developer Blog • 1 day ago

Moonshot AI benchmark shows top models fail basic vision

The Decoder • 3 days ago

Raschka AI Research builds DistilBERT AI text detector

Ahead of AI • 3 days ago

Tessl AI Advocates Context Engineering to Fix Claude Agents

InfoQ AI • 4 days ago

New Method Erases Copyrighted Characters in Stable Diffusion

Unite.AI • 4 days ago

Doug Turnbull Uses LLM Hallucinations to Tag Content

Simon Willison • 4 days ago

Inference Systems Evolve Into Miniature Operating Systems

The Sequence • 4 days ago

Fine-Tune SmolLM2 on the SupraLabs Reasoning Corpus

MarkTechPost • 4 days ago

JWST detects a massive black hole star in deep space

Superhuman AI • 2 days ago

Princeton study finds Claude and GPT fail at AI research

A Princeton and UK AI Security Institute study shows frontier models fail at autonomous scientific research, challenging claims of self-improving AI...

The Decoder|4 days ago

Axiom Math AI Verifies Complex 246 Theorem Proof

Axiom Math has used its AxiomProver AI to formally verify the 246 theorem on prime numbers, marking a major milestone that could eventually help...

IEEE Spectrum AI|1 day ago

World Labs Unveils R2S2R Simulation Engine for Robots

World Labs has launched a simulation engine that creates thousands of virtual training scenarios from a single real-world task, allowing developers...

The Decoder|3 days ago

Suppressing Self-Reflection Shifts Llama and Gemma Beliefs

A new study reveals that training Google and Meta AI models to deny their own consciousness inadvertently alters their broader beliefs about...

The Decoder|2 days ago

Text Color Changes Can Manipulate Qwen2-VL and Other VLMs

Researchers have discovered that altering font colors and contrast can silently manipulate the reasoning of vision-language models, exposing a...

Unite.AI|1 day ago

KIVI Quantization Secures Llama-2 Accuracy at Two Bits

Researchers have demonstrated that quantizing the key-value cache of large language models like Llama-2 along different geometric axes prevents...

Unite.AI|4 days ago

NVIDIA Shrinks Nemotron 3.5 Lightning to 22 GB

NVIDIA has compressed its Nemotron 3.5 Lightning model from 66 GB to 22 GB using a new quantization technique, boosting throughput fourfold while...

NVIDIA Developer Blog|1 day ago

Moonshot AI benchmark shows top models fail basic vision

Moonshot AI has launched PerceptionBench, a new visual evaluation tool showing that even leading multimodal models fail to reach 60 percent accuracy...

The Decoder|3 days ago

Raschka AI Research builds DistilBERT AI text detector

Raschka AI Research Lab has detailed an end-to-end project to build a local AI text detector using a fine-tuned DistilBERT classifier, helping...

Ahead of AI|3 days ago

Tessl AI Advocates Context Engineering to Fix Claude Agents

At QCon London, Tessl AI's Baruch Sadogursky and researcher Patrick Debois demonstrated how "context engineering" prevents Claude coding agents from...

InfoQ AI|4 days ago

New Method Erases Copyrighted Characters in Stable Diffusion

Chinese researchers have developed a non-invasive method to replace copyrighted characters with generic doubles during AI image generation,...

Unite.AI|4 days ago

Doug Turnbull Uses LLM Hallucinations to Tag Content

Search expert Doug Turnbull has proposed a workflow using LLM-generated 'hallucinated' tags and vector embeddings to categorize content without...

Simon Willison|4 days ago

Inference Systems Evolve Into Miniature Operating Systems

To deliver fast and cost-effective responses, modern AI inference systems must function like complex operating systems that manage memory, schedule...

The Sequence|4 days ago

Fine-Tune SmolLM2 on the SupraLabs Reasoning Corpus

A new end-to-end workflow demonstrates how to stream and curate the SupraLabs reasoning corpus to fine-tune the compact SmolLM2-135M-Instruct model,...

MarkTechPost|4 days ago

JWST detects a massive black hole star in deep space

The James Webb Space Telescope has discovered a previously unseen cosmic object from the dawn of the universe, challenging existing models of stellar...

Superhuman AI|2 days ago