White Papers

Executive synthesis of LLMSurgeon: Diagnosing Data Mixture of Large Language Models

Executive synthesis of Physics Is All You Need? A Case Study in Physicist-Supervised AI Development of Scientific Software

Executive synthesis of Reasoning with Sampling: Cutting at Decision Points

Executive synthesis of Unlocking the Working Memory of Large Language Models for Latent Reasoning

Executive synthesis of VideoMLA: Low-Rank Latent KV Cache for Minute-Scale Autoregressive Video Diffusion

Analysis of Recursive Self-Correction via Latent Symbolic Traces (LS-Trace)

Analysis of OmniWorld: Unified Spatiotemporal World Models for Autonomous Generalization

Analysis of SparseHyper-MoE: Dynamic Routing via Hyper-Network Weight Prediction

Executive summary of Efficient Sparse Attention via Dynamic Token Pruning in Multi-Modal Models

Executive summary of Steer-to-Detect: Probing Hidden Representations for Detection of LLM-Generated Texts

Executive summary of Scaling Laws for Neural-Symbolic Integration in Large Language Models

Research synthesis and original whitepaper for Agentic Scaling Laws: Emergent Reasoning in Iterative Loops

Research synthesis and original whitepaper for Ultra-Low Bit Quantization for On-Device Intelligence

Research synthesis and original whitepaper for The Synthetic Data Equilibrium: Avoiding Model Collapse