AI Model2026
A 35B additive ternary-quantized variant of Mach-1 that aims to preserve most capabilities of its BF16 teacher while reducing compute and memory; shows ~95% mean retention across 12 benchmarks and competitive per-task parity on several evaluations.
AI Video Papers2026
Performs real-time, instruction-guided video-to-video editing on streaming input using a 16B autoregressive diffusion model that preserves subject identity and long-term temporal coherence; achieves end-to-end 720p at ≈30 FPS on a single Nvidia B200 GPU. Key features include chunk-wise autoregressive adaptation, Source-Anchored Distribution Matching Distillation (SA-DMD) that reduces diffusion to a two-step generator, and Long-Horizon Autoregressive Distillation to mitigate temporal drift.
Detects and filters spurious token-level teacher supervision in on-policy distillation by estimating input-groundedness and removing high-impact misleading updates, improving OPD on both LLM and VLM benchmarks.
AI Agent Papers2026
Converts sparse trajectory-level rewards into turn-level credit by aggregating token-level teacher–student log-probability gaps and recursively updating a Bayesian belief in log-odds; produces turn-wise reweighting for policy optimization without an extra critic or rollouts.
AI Agent Papers2026
Transfers hierarchical, structured memory from a large teacher agent into small LLM agents to improve tool-use success. Constructs Workflow, Subtask and Function memories with proactive/reactive injection; training-free and validated on multiple tool-use benchmarks.
Describes a 314B-parameter decoder-only Mixture-of-Experts language model that activates 13.2B parameters per token for fine-grained sparsity, long-context (up to 256K) and multi-domain capabilities. Emphasizes GDLA architecture, expert balancing, and multi-teacher distillation.
AI Agent Papers2026
Uses a stronger 'builder' model at inference time to construct executable harnesses that boost weaker target models without parameter updates, mainly by turning unstable reasoning into deterministic code, routing, and strict answer-format enforcement.
AI Video Papers2026
Externalizes persistent scene state into a camera-indexed world bank and designs a long-horizon teacher whose sparse-attention supervision is distilled into a three-step student, enabling responsive, low-latency interactive long-horizon video generation with bounded denoiser context.
AI Video Papers2026
Predicts future video frames conditioned on an observed frame, a language instruction, and a sequence of end-effector poses and gripper states for robot manipulation. Uses per-arm SE(3) geometric encoding (PRoPE-style), a lightweight depth branch, SAM3 masks with a frozen V-JEPA teacher, and distribution-matching distillation for efficient, consistent action-conditioned rollouts.
AI Model2026
A 9B-parameter distillation that transfers chain-of-thought reasoning from Qwen3.8 into the Qwen3.5-9B architecture for single‑GPU deployment; trained on ~70,000 teacher traces, it offers 262k-token context, native function-calling, and improved MMLU performance.
AI Dataset2026
A synthetic, verifiable-first agentic training corpus with 19,072 training traces and 2,135 held-out evaluation rows. Provides per-turn visible reasoning, real sandboxed tool executions, 13 verifiable task families, and NeMo Gym / RL-ready reward contracts for SFT and RL workflows.
Analyzes how on-policy distillation (OPD) transfers teacher LLM capabilities to student models across in-domain shifts, cross-domain transfer, and multi-teacher settings. Key findings: OPD conveys reasoning patterns rather than specific answers, same-origin teacher-student pairs generalize broadly, and multi-teacher combinations induce mixture-dependent tradeoffs.