Converts posed indoor RGB(-D) video into editable, simulation-ready 3D scene graphs by parsing multi-view evidence into per-object bundles, generating complete object assets from that evidence, and placing them with GizmoAct, a VLM policy that refines 9-DoF poses through closed-loop GUI actions.
Turns sparse per-student records into individualized simulators that both reproduce a student’s responses and update them under tutor guidance using pooled LLM pretraining followed by per-student specialization; releases StudentSimEval and reference simulators across chess, L2 writing, and math.
Generates compact keyword sets for both queries and items with LLMs and matches them directly via an inverted index. Uses supervised fine-tuning to align keyword spaces, then alternates GRPO-based reinforcement learning on query- and item-side generators to co-evolve representations and maximize retrieval F1 while staying compatible with keyword-based infrastructure.
Selectively admits dense token-level teacher supervision only after a prompt-level verifier audit, routing prompts that fail the audit to verifier-grounded trajectory supervision instead — reducing harmful updates from confidently wrong teachers and improving teacher GPU utilization.
AI Agent Papers2026
Presents two LLM-based search agents (Iris-mini and Iris-pro) trained by alternating supervised fine-tuning and reinforcement learning against live web search. Key features: web-graph-derived multi-hop tasks with entity abstraction, SFT–RL climbing, inference-time context management, and state-of-the-art open-source benchmark results.
Combines sparse verifier outcomes with dense privileged‑hindsight token scoring to learn an outcome‑calibrated, normalized distribution over complete responses for on‑policy self‑improvement. Key features: sign‑gated guidance (retain/reverse/disable per verifier advantage), profiled trajectory balance with one log‑partition per rollout group, and explicit correction against false‑positive self‑guidance.
AI Model2026
A 2B-class causal LLM packaged as a GGUF for local inference; offers 131072-token native long context, XML-style tool-calling support, and is tuned with SFT + RL + OPD using the UltraData family for stronger code, math and agentic abilities.
Trains end-to-end driving without human trajectory supervision by decoupling perception and action: DriveVFM distills multiple frozen vision foundation models into a single camera backbone, and DriveRL trains a privileged closed-loop RL teacher whose rollouts supervise a camera-only planner, yielding state-of-the-art closed-loop benchmark results.
AI Model2026
Compact causal LLM for on-device assistants, coding agents and long-context tool use — ~2.52B parameters with a 131,072-token context, trained with SFT + RL + OPD and released with its UltraData training corpora and multi-format deployment checkpoints.
AI Dataset2026
Provides ~86K verifiable-reward RL training samples across Math, Knowledge (STEM), Long-Context, and Code for post-training LLMs; each sample includes a ground-truth and verifier-friendly JSONL format for stable reward signals. Note: code tasks require an external sandbox to execute tests.
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.