olmoearth v1.1 cuts compute costs up to 3x
a new family of remote sensing models reduces token sequence length to lower inference costs while matching previous performance.
topic
research
google launches gemini omni for video creation
gemini omni flash can generate and edit videos from text, images, audio, and video inputs, starting with the gemini app and youtube shorts.
google launches gemini for science tools
google introduces experimental ai tools to speed up hypothesis generation, computational discovery, and literature review for researchers.
google's era ai tool writes expert scientific code
google research's empirical research assistance (era) uses gemini to automate and optimize scientific coding, now published in nature and powering the computational discovery prototype.
light-matter particles could speed up ai chips
penn researchers created exciton-polaritons that switch signals with tiny energy, potentially enabling faster, more efficient photonic ai computing.
counterfactual paths cut credit assignment noise in llm reasoning
a new method uses counterfactual trajectory comparisons to turn sparse terminal rewards into step-level signals, stabilizing reinforcement learning for multi-step llm reasoning.
adversarial action removal breaks self-play rl agents
attackers who delete legal actions before an agent decides cause severe and lasting damage across multiple games and algorithms.
signmuon: 1-bit matrix-aware optimizer for distributed training
signmuon combines 1-bit sign communication with muon's matrix-aware updates to reduce distributed training bottlenecks.
skim speeds up web agents with speculative execution
skim uses site-specific templates to skip heavy ai steps for most web tasks, falling back to full agents only when needed.
agentwall adds runtime safety for local ai agents
agentwall intercepts agent actions before execution, enforcing declarative safety policies to prevent harmful operations on local machines.
stopping local ai agents early saves energy
agentstop reduces energy use in local ai agents by predicting task failure early, cutting token waste and battery drain on consumer devices.
risk-aware policy learning for offline bandits
a new method learns safe decision rules from logged data using general risk measures like cvar, with strong theoretical guarantees.
six months of llm progress in five minutes
a lightning talk summary of major llm developments from november 2025 to may 2026, including coding agent breakthroughs and open-weight model advances.
sandboxaq puts drug discovery ai into claude
sandboxaq integrates its physics-grounded quantitative models into anthropic's claude, letting researchers run complex simulations through plain language prompts.
latent bias in llms can flip decisions
instruction-tuned language models show fair outputs but retain biased internal representations that can reverse decisions when activated.
teamtr fine-tunes multi-agent llms with trust regions
a trust-region method for fine-tuning multi-agent llm teams avoids compounding errors from stale rollouts, outperforming baselines by 7.1%.
open agent leaderboard compares full ai systems
a new open benchmark evaluates complete agent systems, not just models, across six diverse tasks to measure generality, quality, and cost.
deepslide: ai for full presentation delivery
deepslide is a multi-agent system that helps prepare entire presentations, from planning and slide creation to rehearsal support.