Predicting Privileged Information Effectiveness for On-Policy Self-Distillation
Paper • 2610.22512 • Published • 10
🤝 Open to Collab
Urro
urroxyz
AI & ML interests
computational linguistics major 🤖🔎🔠
i am autistic. if i come off rude, i probably didn't mean to. please feel free to ask me for clarification.
Recent Activity
updated a collection 9 days ago
WTF GENIUS PAPERS upvoted a paper 9 days ago
Predicting Privileged Information Effectiveness for On-Policy Self-Distillation updated a collection 9 days ago
WTF GENIUS PAPERSOrganizations
upvoted 10
papers 9 days ago
TokenGate: Adapting LLMs for Fast and Memory-Efficient Prefill with Learned Token Pruning
Paper • 2610.26754 • Published • 11
Good Judgment, Bad Advice: Decision-Model Controllers Raise Agent Cost Without Improving Success
Paper • 2610.24649 • Published • 16
On the Tip of the Tongue: First-Token Cueing as a Behavioral Probe of Recall Failures in LLMs
Paper • 2610.23400 • Published • 8
Beyond Compilation: Evaluating Faithful Natural-Language-to-Lean Statement Formalization
Paper • 2606.31002 • Published • 9
On the Fragility of Tasks under Composition: Implicit Budgeting in Language Models
Paper • 2610.27190 • Published • 11
Self-Credit: Credit Assignment for Long-Horizon Agentic RL via Intra-group Rescaling
Paper • 2610.28388 • Published • 13
Large Flow Language Models
Paper • 2610.28559 • Published • 24
Breaking Diffusion: Parallel Reasoning with Draft Refinement
Paper • 2610.26542 • Published • 24
PAR: Progress-Aligned Recurrence for Length Generalization in State Tracking
Paper • 2610.27628 • Published • 169
upvoted 6
papers 11 days ago
Sparse Autoencoders Are a Shared Coordinate System for Symbolic Matching
Paper • 2610.19636 • Published • 8
Lexical Reference Frames: Interpreting LM Residual Streams in Native Semantics
Paper • 2610.22169 • Published • 51
Tree-Aware Posterior Distillation for Long-Horizon Agentic Reinforcement Learning
Paper • 2610.23181 • Published • 57
Embeddings Are Overpriced? A Frozen 16 MB Vocabulary for Small and Ternary Language Models
Paper • 2610.22428 • Published • 21
OPMD: On-Policy Mechanism Distillation Beyond Output Supervision
Paper • 2610.14022 • Published • 20
Towards Looped Models Done Right, Part I: Topology, Input Injection, Recurrent-State Design
Paper • 2610.23036 • Published • 19
upvoted 4
papers 12 days ago
ProbabilityBench: Model Evaluation and Calibration with Exact Ground-Truth Uncertainty
Paper • 2610.17016 • Published • 14
FreeSolo: Linear-time Sparse Attention with Zero-Shot Length Extrapolation
Paper • 2610.17191 • Published • 15
History Replays Itself: Unleashing Long-Context Capability in Linear Attention
Paper • 2610.15715 • Published • 26
Learning to Read Out: Unembedding Dynamics in Language Model Pretraining
Paper • 2610.21025 • Published • 10