The Archive

Higher-order optimization algorithms such as Shampoo have been effectively applied in neural network training for at least a decade. These methods have achieved... Higher-order optimization algorithms such as Shampoo have been effectively applied in neural network training for at least a decade. These methods have achieved significant success more recently when applied to leading LLMs. In particular, Muon (MomentUm Orthogonalized by Newton-Schulz) was used to train some of today’s best open source models, including Kimi K2 and GLM-5. Source

User guide for non-technical knowledge work using Claude Cowork, emphasizing multi-step workflows without coding.

Cursor was on track to close a $2 billion funding round this week but chose to halt discussions after SpaceX offered a $10 billion "collaboration fee" and a path to a $60 billion acquisition.

Benchmark analysis shows Qwen3.6-27B dense vs. Qwen3.6-35B-A3B MoE gap narrowing; MoE gains +4.9pp on SWE-bench coding but dense leads overall.

Claude Code is being moved behind a $100/mo paywall. Would you pay that for an AI coding tool?

Shopify CTO discusses 2026 AI adoption roadmap, unlimited Claude Opus 4.6 token budget, and internal tools (Tangle, Tangent, SimGym).

User reports Qwen3 TTS running locally in real-time with expressive output, integrated into Persona Engine lip-sync pipeline.

Google's newest TPUs are faster and cheaper than the previous versions. But the company is still embracing Nvidia in its cloud — for now.

Untenable demand has Anthropic exploring new approaches to rationing its service.

Reddit user reports Claude Opus 4.7 performs worse than 4.6 for their use case; expresses frustration with degradation.

SpeechParaling-Bench expands paralinguistic feature evaluation for Large Audio-Language Models from 50 to 100+ features across 1,000+ English-Chinese speech queries.

Parallel-SFT improves zero-shot cross-programming-language transfer for code RL by enabling skills learned in high-resource languages like Python to transfer to lower-resource languages.

AVISE is an open-source modular framework for identifying vulnerabilities and evaluating security of AI systems using extended Red Queen attack theory.

FedSIR proposes spectral client identification and relabeling to mitigate label noise in federated learning across distributed clients.

CS-ARM-BN addresses batch effects in biomedical imaging via meta-learning adaptation using in-context control samples to improve domain generalization.

Global Sentinel-1 SAR time series dataset monitors offshore wind infrastructure deployment and operations from 2016Q1 onward at high temporal resolution.

Stream-CQSA avoids out-of-memory failures in long-context LLMs by decomposing quadratic attention via cyclic quorum set theory into independent subsequence computations.

Study shows Transformers, LSTMs, RNNs, and word embeddings converge on period-T periodic features for number representation; proves Fourier sparsity is necessary but insufficient for mod-T separability.

ParetoSlider enables multi-objective RL post-training for diffusion models with inference-time control over conflicting reward criteria without early scalarization.

Ace is the first robot that can beat the best human players while following the official rules of table tennis. | Image: Sony AI Humans have been building ping-pong playing robots for decades, such as Omron's FOREPHUS that challenged amateur competitors at CES 2017. What sets Ace apart from the rest is that the robot, which was developed by Sony's AI division, is the first that can hold its own against top-ranked human players and occasionally even beat them in matches that follow the official rules of the International Table Tennis Federation (ITTF). AI is already capable of besting humans a...

RoboGrid framework reveals LLMs maintain surface syntax but fail semantic/behavioral adherence to context-free grammars under recursion depth and complexity stress-tests.

User shares comprehensive cheat sheet for Claude Code covering shortcuts, workflows, and MCP setup.

OMIBench evaluates vision-language models on multi-image Olympiad-level reasoning across biology, chemistry, mathematics, and physics.

Principal-agent framework reconceptualizes AI alignment as structural governance problem across objectives, information, and multiple stakeholders.

Google brings Gemini-powered 'auto browse' capabilities to Chrome for enterprise users, letting workers automate tasks like research, data entry, and more.

LEXIS reconstructs 3D human-object interactions from RGB images using dense proximity field representations.

Gauge-equivariant graph neural networks embed non-Abelian symmetries into message passing for lattice gauge theory problems.

Hybrid architecture extends LLMs with RDF/OWL ontology layer for structured, verifiable reasoning and automated knowledge graph construction.

Clinical LLM evaluation reveals alignment gaps with physician communication standards in empathy, readability, and emotional affect.

Cognitively-inspired working memory constraints and fixed-width attention improve Transformer performance on low-data grammar tasks.