Research papers, read to you in five minutes.
Research papers, read to you in five minutes.

GraphForge tackles the real-files-vs-reliable-grading tradeoff in agent training data: it crawls actual documents, builds a cross-file evidence graph, then compiles rubrics whose criteria point back to specific files. Fine-tuning Qwen3 27B on just 2,169 trajectories lifts GDPVal by 65.7 Elo across three unseen scaffolds.

Diffusion · Added Oct 3

Adaptive Reward Routing: Dynamic Multi-Reward Optimization for Joint Audio-Video Diffusion via Forward-Process RL

Multi-reward RL on joint audio-video diffusion models silently kills weak objectives like A/V sync when gradient-based weighters zero them out, and static modality routing goes stale as cross-modal layers shift during fine-tuning. Adaptive Reward Routing re-routes gradients each round and floors user priors, cutting DeSync from 0.604 to 0.341.

Agents · Added Oct 3

Beyond Memory: Harnessing Long-Horizon Agents with Explicit Belief States

PoS tackles long-horizon agents acting on stale or self-contradictory world state by maintaining an explicit entity-state-relation belief, auditing updates, and detecting "belief trapping" via a sliding-window health score. Wins all 12 benchmark cells with up to 37.89% gains on root-cause analysis, but costs 5× the tokens.

Agents · Added Oct 3

OneStreamer: Unifying Perception, Memory, and Proactive Response in Streaming Video Interaction

Streaming video models forget: a 16-frame window evicts the second-7 event before the minute-2 question arrives, but keeping full history wrecks present perception. The fix has a 4B model write itself timestamped text notes mid-stream, cutting context from 62k to 4.3k tokens while beating 11B baselines.

Evaluation · Added Oct 3

On-Policy or Off-Policy Learning? A Systematic Study of Distillation Dynamics

On-policy distillation's reputation for less forgetting, sparser updates, and better generalization mostly dissolves under controlled testing: KL direction and learning rate explain the gains, with teacher-rollout training matching student-rollout within 1 point accuracy at a fraction of the compute.

Never fall behind the literature again.

Free account. Follow topics, build your queue.

Sign up free

Agents · Added Oct 4

GraphForge: Training Working Agents with Graph-Anchored Workspace Synthesis

GraphForge tackles the real-files-vs-reliable-grading tradeoff in agent training data: it crawls actual documents, builds a cross-file evidence graph, then compiles rubrics whose criteria point back to specific files. Fine-tuning Qwen3 27B on just 2,169 trajectories lifts GDPVal by 65.7 Elo across three unseen scaffolds.

Diffusion · Added Oct 3

Adaptive Reward Routing: Dynamic Multi-Reward Optimization for Joint Audio-Video Diffusion via Forward-Process RL

Multi-reward RL on joint audio-video diffusion models silently kills weak objectives like A/V sync when gradient-based weighters zero them out, and static modality routing goes stale as cross-modal layers shift during fine-tuning. Adaptive Reward Routing re-routes gradients each round and floors user priors, cutting DeSync from 0.604 to 0.341.

Agents · Added Oct 3

Beyond Memory: Harnessing Long-Horizon Agents with Explicit Belief States

PoS tackles long-horizon agents acting on stale or self-contradictory world state by maintaining an explicit entity-state-relation belief, auditing updates, and detecting "belief trapping" via a sliding-window health score. Wins all 12 benchmark cells with up to 37.89% gains on root-cause analysis, but costs 5× the tokens.

Agents · Added Oct 3

OneStreamer: Unifying Perception, Memory, and Proactive Response in Streaming Video Interaction

Streaming video models forget: a 16-frame window evicts the second-7 event before the minute-2 question arrives, but keeping full history wrecks present perception. The fix has a 4B model write itself timestamped text notes mid-stream, cutting context from 62k to 4.3k tokens while beating 11B baselines.

Evaluation · Added Oct 3

On-Policy or Off-Policy Learning? A Systematic Study of Distillation Dynamics

On-policy distillation's reputation for less forgetting, sparser updates, and better generalization mostly dissolves under controlled testing: KL direction and learning rate explain the gains, with teacher-rollout training matching student-rollout within 1 point accuracy at a fraction of the compute.

Never fall behind the literature again.

Free account. Follow topics, build your queue.

Sign up free

Browse topics

Computer Vision 149

Video Generation 66

Reinforcement Learning 106

Research papers, read to you in five minutes.Research papers, read to you in five minutes.

Research papers, read to you in five minutes.
Research papers, read to you in five minutes.