PhysicsMind: Sim and Real Mechanics Benchmarking for Physical Reasoning and Prediction in Foundational VLMs and World Models
CW Mak, G Zhu, et al. arXiv preprint · 2026
DeciLix Lab. Decision Intelligence and Lattice Systems.
A research collective working on agents, large language models, recommender systems, and finance. We ship what we publish.
Agent systems hold up outside the lab, or they do not hold up at all. We publish in the open, ship to production, and learn from what breaks.
Selected work
All papers (30) →What we ship
All projects →Benchmark
SkillAudit
Skill-centered assessment for agent skills across utility, efficiency and cost, and safety, backed by sandboxed execution evidence.
Benchmark
PhysicsMind
A simulation-and-real-world benchmark for testing physical reasoning and prediction in vision-language and world models.
Research
Latent Thinking Optimization
Code for supervising and improving latent reasoning by treating hidden-state correctness signals as a latent reward model.
Research
LLMPopcorn
An LLM-assisted pipeline for generating and evaluating titles, cover prompts, and short-video prompts designed for audience appeal.
Benchmark
MMPCBench
A benchmark for measuring how multimodal language models reconstruct missing product text or imagery and support recommendation.
Open source
MetaSR
The official implementation of a meta-learning framework that bridges next-item prediction and masked-language modeling for recommendation.
Latest insights
All insights →Work in the open. Ship in the real world.
We are hiring researchers, engineers, and collaborators. We also welcome paper submissions and partnership proposals.