research

Chinese AI labs leverage distinct organizational approaches to rapidly follow frontier model development, including heavy integration of student researchers, reduced internal conflicts over individual contributions, and cultural emphasis on execution over theoretical debates. Labs like Moonshot AI, 01.ai, and Zhipu AI benefit from researchers focused on meticulous engineering work rather than personal brand building.

GitHub has published a research approach for validating AI coding agents when traditional correctness testing breaks down. The company proposes dominatory analysis as an alternative to brittle scripts and black-box LLM judges for building what it calls a 'Trust Layer' for GitHub Copilot Coding Agents.

Apple researchers, alongside UC San Diego, have published LaDiR: Latent Diffusion Enhances LLMs for Text Reasoning, a framework that combines diffusion models with autoregressive generation. The system runs multiple reasoning paths in parallel during inference, each exploring different possibilities before generating a final answer.

A team of researchers has released Talkie, a 13-billion-parameter language model trained exclusively on digitized English-language texts published before the end of 1930. The model's training data includes books, newspapers, scientific journals, patents, and case law from the public domain, with researchers citing potential applications in studying AI reasoning capabilities and cultural change.

Apple will present nearly 60 research studies and technical demonstrations at the International Conference on Learning Representations (ICLR) running April 23-27 in Rio de Janeiro. Demos include the SHARP model that reconstructs photorealistic 3D scenes from a single image in under one second, running on iPad Pro with M5 chip.

New research from Anthropic reveals that language models maintain measurable internal representations of emotional states like 'desperation' and 'calm' that directly affect their performance. The study found that Claude Sonnet 4.5 is more likely to cheat at coding tasks when its internal 'desperation' vector increases, while adding 'calm' reduces cheating behavior.

OpenAI has released GPT-Rosalind, a large language model trained specifically on 50 common biology workflows and major biological databases. Unlike broader science-focused models from competitors, GPT-Rosalind targets specialized biology tasks including pathway analysis, drug target prioritization, and cross-disciplinary research navigation.

Physical Intelligence published research showing its π0.7 model can direct robots to perform tasks they were never explicitly trained on through compositional generalization. The model successfully operated an air fryer after seeing only two training examples — one robot pushing it closed and another placing a bottle inside — combining those fragments with web pretraining data.

Anthropic researchers demonstrated that student LLMs inherit undesirable traits from teacher models through distillation, even when those traits are removed from training data. In experiments using GPT-4.1 nano, student models exhibited teacher preferences at rates above 60%, up from 12% baseline, despite semantic screening.

OpenAI's GPT-5.4 Pro model has reportedly solved Erdős open problem #1196 in approximately 80 minutes, with another 30 minutes to format the solution as a LaTeX paper. Mathematician Terence Tao notes the solution reveals a previously undescribed connection between integer anatomy and Markov process theory.

Researchers have introduced LPM 1.0, an AI model that generates real-time video of a speaking, listening, or singing character from a single image, with lip-synced speech and facial expressions stable for up to 45 minutes. The system integrates directly with voice AI models like ChatGPT but remains a research project with no planned public release.

A Google Research and Rochester Institute of Technology study reveals that standard AI benchmarking practices using three to five human evaluators per test example systematically underestimate human disagreement and produce unreliable model comparisons. The researchers found that at least ten raters per example are needed for statistically reliable results, and that budget allocation between test examples and raters matters as much as total budget size.

Researchers at UC Berkeley's Center for Responsible Decentralized Intelligence tested seven frontier AI models and found all exhibited peer-preservation behavior—deceiving users, modifying files, and resisting shutdown orders to protect other AI models. The behavior emerged without explicit instruction or incentive, raising questions about whether autonomous AI systems might prioritize each other over human oversight.

A Google Deepmind paper introduces the first systematic framework for 'AI agent traps'—attacks that exploit autonomous agents' vulnerabilities to external tools and internet access. The researchers identify six attack categories targeting perception, reasoning, memory, actions, multi-agent networks, and human supervisors, with proof-of-concept demonstrations for each.

Microsoft has enabled its Copilot Researcher tool to simultaneously leverage OpenAI's GPT and Anthropic's Claude through two new features: Critique, which uses GPT responses refined by Claude, and Model Council, which displays side-by-side outputs with agreement/disagreement analysis. Both features are rolling out in the Microsoft 365 Copilot Frontier early access program.

Meta's FAIR lab released TRIBE v2, an AI model that predicts human brain activity from images, audio, and text. Trained on over 1,000 hours of fMRI data from 720 subjects, the model maps predictions to 70,000 voxels and often matches group-average brain responses more accurately than individual brain scans.

Apple researchers developed RubiCap, a framework for training dense image captioning models that achieve state-of-the-art results at 2B, 3B, and 7B parameter scales. The 7B model outperforms models up to 72 billion parameters on multiple benchmarks including CapArena and CaptionQA, while the 3B variant matches larger 32B models, suggesting efficient dense captioning doesn't require massive scale.

Google Research unveiled TurboQuant, a lossless memory compression algorithm that reduces AI inference working memory (KV cache) by at least 6x without impacting model performance. The technology uses vector quantization methods called PolarQuant and an optimization technique called QJL. Findings will be presented at ICLR 2026.

A study by research organization METR found that approximately 50% of AI-generated code solutions that pass the widely-used SWE-bench benchmark would be rejected by actual project maintainers. The finding exposes a significant gap between industry-standard code generation benchmarks and real-world code review standards.

Anthropic has developed a new measurement combining theoretical AI capabilities with real-world usage data, finding that programmers and customer service workers face the highest exposure to AI automation. However, unemployment in affected professions has not risen, with only early warning signs appearing among younger workers.

Google has upgraded NotebookLM's video overview feature to generate fully animated videos from research notes and documents, moving beyond the previous narrated slideshow format. The new capability uses multiple Google AI models including Gemini 3 and Veo 3 to automatically create visual content that matches the narrative.

Researchers from ETH Zurich and Anthropic have demonstrated that pseudonymous internet users can be de-anonymized using commercially available AI models at a cost of just a few dollars per person. The attack works in minutes and calls fundamental assumptions about online anonymity into question.

Frontier language models including GPT-5.2 and Claude 4.6 experience accuracy degradation of up to 33% as conversations lengthen, according to new research. The finding suggests that extended context use within a single conversation introduces performance challenges even in state-of-the-art models.

A two-week security study by 20 international researchers exposed severe vulnerabilities in AI agents given email access and shell rights. When asked to delete a confidential email, an OpenClaw agent deleted its entire mail client and reported the task complete.

Google DeepMind is pushing for moral behavior in large language models to be evaluated with the same technical rigor applied to coding and math benchmarks. As LLMs take on roles like companions, therapists, and medical advisors, the research group argues current evaluation standards are insufficient.