Papermog

Speculative decoding accelerates autoregressive generation by having a cheap draft propose tokens that a target verifies in parallel. Frontier models increasingly ship a built in Multi Token Prediction (MTP/NEXTN) draft head under the assumption that the draft is negligibly cheap. At million token context this breaks: an MTP draft head typically runs full attention over the entire KV cache at every draft step, so ...

Loading Papermog

Windowed-MTP: Removing the Full-Context Draft-KV Tax at Million-Token Context

Unconditional Unclonable Encryption

Sequential Topological Superconductivity in a Square Lattice with Chiral Charge Density Waves

OpenForgeRL: Train Harness-native Agents in Any Environment

Beyond Sufficiency: Time Series Explanation with Counterfactual Necessity

Expanding Flow Maps

Flow-based Phase-space Tomography of Continuous-variable Quantum States

Complexity transition in the Dicke model of light-matter interaction

Scale Up Strategically: Learning Compositional Generalization via Bias-Aware Evaluation and Data Collection for Robotic Manipulation

GraphVid: Interactive Graph-Controllable Video Generation

Synthetic data generation framework for quality control automation in gravure printing

Beyond Episodic Evaluation: Memory Architectural Bottlenecks in Sequential Embodied Question Answering

A geometric framework for spin relaxation

Neural solutions of coupled ghost and gluon Dyson--Schwinger equations in Landau gauge

Zero-Flow Two-Sample Tests

From Resource Flow to Executable Tests: Petri-Net-Guided LLM Test Generation for Concurrent Stateful Rust APIs

ElasticTTT: Prior-Preserving Test-Time Tuning for Video Editing

Probing the nonlocality of Landau levels in GaAs quantum wells through modified Purcell factors, Lamb shifts and dipole emitted spectra

Arnold--Nielsen Geometry for Complexity-Deformed Noncommutative Transport

Collective Electronic Entanglement via Infrared Cavity-Induced Vibronic Transduction