Evaluates a weak teacher's RL-induced policy shift on the student's own rollouts and amplifies verifier-supported updates so stronger models can learn from weaker supervisors and surpass them. It rescales only verifier-supported policy-gradient components to preserve optimization fixed points while accelerating learning, reducing student updates versus standard RL or distillation.
Generates long-form, text-controlled music with explicit arrangement and planning. Uses a 50 Hz single-codebook tokenizer, a flow-matching diffusion Transformer to predict VAE latents, and an MoE autoregressor with ABC‑CoT planning to produce 48 kHz audio up to 5m30s.
Reframes Policy Mirror Descent for autoregressive LLM generation into a critic-free, trajectory-level objective using the Bellman equations to avoid per-token value estimation. Derives a practical token-level loss with a smoothed mismatch-correction weight and shows improved math-reasoning accuracy on benchmarks.
AI Agent Papers2026
Turns past discovery traces into replayable simulators so alternative exploration policies can be evaluated offline ('dreaming'), enabling fast, low-cost meta-level policy improvement for agent-driven discovery across coding, optimization, and GPU-kernel tasks.
AI Model2026
Explains how Jev turns input state into typed decisions and probabilities without generating text. Introduces parallel sampling and RLCD training, with workflow evaluations and caveats for software automation.
Introduces physical rank consistency (PRC) and ActionPiece, a learned action tokenizer that preserves local physical relationships via joint supervision of representation learning and quantization, improving autoregressive vision-language-action policy performance across multiple benchmarks.
AI Agent Papers2026
Transforms scientific code repositories into executable, agent-learnable environments that support task generation, execution, and scientific verification. Agent-guided repository transformation produces verified interaction trajectories used to train the PhAI-IDE model family. Intended for research on agent learning, scientific-code repair, and training RL/SFT models.
Analyzes a failure mode in PPO critics called “Value Flattening” and proposes SP³O (Sparse PPO), which supervises a small set of well-separated states per response to mitigate flattened value estimates and improve policy learning for LLM alignment and reasoning tasks.
AI Dataset2026
Provides a complete benchmark and training release for native audio‑visual dialogue: 2,800 synthesized audio‑visual dialogues, tiered rubrics, scoring code, ~28GB of 1080p media, and an RL reward recipe to evaluate and train omni models that take video+audio and return text.
AI Model2026
Fast, non-autoregressive decision engine that answers typed questions (choice/score/noul) over text or JSON states with calibrated probabilities and confidences in a single forward pass. Suited for routing, triage and moderation workflows; includes a Router to pick checkpoints per request.
AI Video Papers2026
Zhenchen Tang, Yang Li +7New Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, School of Artificial Intelligence, University of Chinese Academy of Sciences +2
Introduces a rubric-based approach for video reward modeling that generates explicit, query-adaptive evaluation criteria before scoring to reduce scalar drift; proposes RGPO, a two-stage training (seed warm-up + joint optimization) that achieves strong pointwise and pairwise evaluation with high data efficiency.
AI Model2026
A non-autoregressive, multilingual System‑1 decision model that answers typed questions (choice/score/noul) over text or JSON in a single forward pass with probability distributions. Key differences: calibrated-probability design, sub-35ms GPU latency, and built-in multilingual routing; requires per-workflow recalibration and validation on long or low-resource languages.