AI research
Source reports covering AI research. Stories will appear when analysis and community reactions are ready.
65 reports · page 2 of 2
Connected topics are temporarily unavailable. You can still browse stories below.
Source reports
The AI revolution in math has arrived
news
24 min read
In July 2025, AI models successfully solved five out of six problems at the International Mathematical Olympiad, surprising mathematicians with their rapid advancement. Despite these impressive results, the impact of AI on research mathematics remains uncertain.
Further human + AI + proof assistant work on Knuth's "Claude Cycles" problem
research
1 min read
Claude found an odd-m construction for the open Hamiltonian decomposition problem, impressing Prof. Donald Knuth, who titled the resulting paper "Claude’s Cycles." The updated research reveals that for the base case m=3, there are 11,502 Hamiltonian cycles, with 996 generalizing to all odd-m, and 760 valid "Claude-like" decompositions identified.
Anthropic Subprocessor Changes
tool
1 min read
Anthropic's Trust Center focuses on AI safety and research, emphasizing transparency and secure practices in AI development. The initiative aims to support the safe transition to transformative AI technologies.
Epoch confirms GPT5.4 Pro solved a frontier math open problem
research
5 min read
A Ramsey-style problem on hypergraphs has been solved by Kevin Barreto and Liam Price using GPT-5.4 Pro. The solution has been confirmed by Will Brian and will be published, along with a transcript of the original conversation.
Attention Residuals
tool
3 min read
Attention Residuals (AttnRes) serves as a drop-in replacement for standard residual connections in Transformers, allowing each layer to selectively aggregate earlier representations. It includes two variants: Full AttnRes, where each layer attends over all previous outputs, and Block AttnRes, which groups layers into blocks to reduce memory usage from O(Ld) to O(Nd).
Measuring progress toward AGI: A cognitive framework
research
2 min read
Artificial General Intelligence (AGI) can accelerate scientific discovery and address significant global challenges. Evaluating progress toward AGI requires diverse methods and approaches, with cognitive science offering valuable tools for assessment.
LLM Architecture Gallery
tool
8 min read
The LLM Architecture Gallery compiles architecture figures and fact sheets from significant LLM comparisons. Users can enlarge figures and navigate to corresponding sections using model titles.
Billion-Parameter Theories
research
10 min read
Billion-parameter theories aim to explain complex phenomena in the universe using concise mathematical formulations. Historical explanations of natural events transitioned from mystical interpretations to scientific inquiry with succinct equations like F=ma and E=mc².
Something is afoot in the land of Qwen
news
4 min read
Alibaba's Qwen team has released the Qwen 3.5 family of open weight models. Junyang Lin, the lead researcher, announced his departure from the team via Twitter.
Productivity gains from AI coding assistants haven’t budged past 10% – survey
research
5 min read
93% of developers utilize AI coding assistants, according to research from Laura Tacho, CTO at DX. Despite high usage rates, overall productivity among developers remains at only 10%.
Gemini 3 Deep Think
news
1 min read
Gemini 3 Deep Think has been upgraded to enhance its specialized reasoning capabilities for addressing modern science, research, and engineering challenges. The Wang Lab at Duke University is utilizing this technology to design new semiconductor materials.
Mathematicians issue a major challenge to AI—show us your work
news
3 min read
Top mathematicians have issued a challenge to AI systems to solve a set of unsolved mathematical problems within a week, as part of an initiative called “First Proof.” These problems are newly formulated and are not present in any large language model's training data.
First Proof
research
1 min read
A set of ten research-level mathematics questions has been created to evaluate the capabilities of current AI systems in providing correct answers. The answers to these questions are known to the authors but will remain encrypted temporarily.
In a study, AI model OpenScholar synthesizes scientific research and cites sources as accurately as human experts
research
4 min read
OpenScholar synthesizes scientific research and accurately cites sources, achieving a level of accuracy comparable to human experts. In a study, researchers found that the AI model significantly reduces the issue of hallucination seen in other models like GPT-4o, which fabricated 78-90% of its outputs.
FORTH? Really!?
opinion
3 min read
FORTH and associative/applicative languages may be more suitable for transformer architectures than traditional top-down problem-solving methods. Generating outputs before their constituent parts could enhance the effectiveness of large language models.