acceptodds: the papers of ICLR 2027
Abstract
The safe for work edition of acceptodds: the submissions, without the markets. Search them, browse them on a map laid out by what they are about, follow the ones you care about, keep reading lists with your group, and get recommendations from what you read. There are no odds, no bets and no reputation here. acceptodds is not affiliated with any organisation or institution. acceptodds is an open source project, open to contributions, at github.com/PotentialHedge/acceptodds-sfw.1 Follow for updates on X.
[PDF]Do Vision Transformers Need All-to-All Attention? Global Communication Through Elastic Learned Cores▲233.33
[PDF]Knowledge Capacity Beyond Independent Facts: Shared Structure, Abstraction and Frequency▼2373.33
[PDF]Erase to Forge, Verify to Defend: Diffusion-based Forgery Attack and Self-verifiable Defense Mechanism▲6.5k3.33
[PDF]Seeing Action-Critical Evidence: Token Pruning with Internal Action Priors for Vision-Language-Action Models▲233.33
[PDF]From Task-Oriented Agents to Personalized Teammates: Understanding Continual Adaptation in Cooperative Games▲1.7k3.33
[PDF]From Fixed Grids to Moving Particles: A Transferable Latent Operator for Fluid Dynamics▲1.7k3.33
[PDF]Sparrow: Sparse Rollout for Stable and Efficient Long-context RL of Large Language Models▲233.33
[PDF]Weighting Schedules Govern What and When Score-Based Generative Models Learn from Multimodal Data▲1.7k3.33
[PDF]Learn Feasibility Once, Optimize All Objectives: Derivative-Free Diffusion Models for Chance-Constrained Programming▲233.33
[PDF]REVEALING AFTER OVERWRITING: AN EXPONENTIAL POMDP OPE LOWER BOUND UNDER HISTORY-DEPENDENT LOGGING▲233.33
[PDF]Quantization-conditioned Normalizing Flows for Multi-granularity Graph Anomaly Detection▲233.33
[PDF]Draft in Parallel, Condition Through Depth: Adjacent Causal Injection for Speculative Decoding▲233.33
[PDF]Alignment and Convergence Analyses of Gradient Flow in Two-Layer Convolutional Neural Networks▲6.5k3.33
1 If this project is interesting to you, reach out at [email protected].