ArXiv link for STEPS: Selective On-Policy Self-Distillation for Reasoning

Few classical games have been regarded as such significant benchmarks of artificial intelligence as to have justified training costs in the millions of dollars. Among these, Stratego -- a board wargam...

ArXiv link for Revisiting scaling laws for reward optimization

Ataraxos, an AI for the board wargame Stratego, establishes a design pattern for reinforcement learning and search that is effective under large amounts of hidden information, a longstanding desiderat...

ArXiv link for Latent Generative Solvers for Generalizable Long-Term Physics Simulation

Julia 1 is our compact decision model. Explore the research, evaluation results, limitations, and model repository.

ArXiv link for MemCalib: Benchmarking and Optimizing Memory Use in LLM Agents

Yesterday was Grok 4.7 (pelicans) and MiMo v2.6 Flash/Pro (more pelicans). Today Anthropic released Claude Opus 5.5, and around an hour later OpenAI released GPT-6 Sol and GPT-6 Luna. It’s …

ArXiv link for RAVE: Re-Allocating Visual Attention in Large Multimodal Models

ArXiv link for MemoryArena: Benchmarking Agent Memory in Interdependent Multi-Session Agentic Tasks

Evaluates typed decisions (choice, score, noul) over 100+ languages in a single forward pass with calibrated probabilities. Outperforms TypeSafe Jev.

ArXiv link for Generative Query Suggestion via Intent Coverage and Query-Level Credit Assignment

Large language models sometimes behave in ways resembling human emotional responses, and recent work has identified internal representations that may explain this. We ask whether LLMs represent pain d...

ArXiv link for Ambient Dataloops: Generative Models for Dataset Refinement

ArXiv link for Theories of Mind as Domain-Specific Languages of Thought

ArXiv link for The "Curse of Knowledge" in LLM Query Simulation: Concept Provenance for Tracing Answer-Side Intrusion

ArXiv link for Workspace Models: Lightweight Robotic Memory via Saliency-Driven Supervision

...or how to make Qwen learn query optimization via agentic reinforcement learning

ArXiv link for Self Improvement via Fast Tree-search

ArXiv link for Are LLMs Good Financial User Simulators? A Preliminary Study

ArXiv link for Opt-In Art: Learning Art Styles Only from Few Examples

ArXiv link for Does Video Memory Use What It Retrieves? A Causal Audit of Memory Specificity

ArXiv link for Forty Shades of Blue: Quality-Diversity Alignment via Mode-Conditioned Reinforcement Learning

When providing forecasted probabilities with a predictive model, the ideal model offers perfect calibration: the true probability of the outcome (i.e., the probability that $Y=1$) exactly matches the ...

The policy below is playing Craftax.

In recent years, a number of prominent computer scientists, along with academics in fields such as philosophy and physics, have lent credence to the notion that machines may one day become as large as...