CompLearn @ ICML 2026

124 papers

Top Research Directions

Research directions most represented by CompLearn @ ICML 2026 papers.

Activation Patching for Mechanistic Interpretability 5
Resolving Superposition with Sparse Autoencoders 4
The Compositional Generalization Gap in Neural Networks 4
Scaling Reasoning Performance via Structural Scaffolding 3
Cascading Hallucinations in Multi Agent Systems 3
Input Adaptive Compute Allocation for Language Models 3
Latent State Communication in Neural Pipelines and Agents 2
Testing Compositional Generalization through Systematic Data Splits 2
Overcoming Interaction Primitive Constraints in Autonomous Agents 2
The Performance Cliff of Multi-Constraint Instruction Following 2
Understanding Feature Superposition in Deep Neural Networks 2
The Gap Between Neural Expressivity and Practical Learnability 2
The Computational Depth Bottleneck in Transformers 2
Activation Steering for Inference Time Model Control 2
Causal Interventions on Neural Representations 2

Top Authors

AuthorPapers
Tae-Hyun Oh4
Arnas Uselis2
Asad Aali2
Ayushi Mehrotra2
Cihang Xie2
Han Liu2
Haoqin Tu2
Haozhou Gao2
Hardy Chen2
Heejeong Nam2
Hong-Yu Chen2
Jacob Andreas2
Jerry Yao-Chieh Hu2
Jiyeon Son2
Jon Kleinberg2
Juncheng Wu2
Kenny Peng2
Letian Zhang2
Martha Lewis2
Martin Pawelczyk2

Papers

Title
Spatially Stable GUI Grounding via Zoom Consistency Loss
Stop Probing, Start Coding: Why Linear Probes and Sparse Autoencoders Fail at Compositional Generalisation
Structure over Pixels: Learning Variable-Length Visual Programs
Successor Re-grounding Audits Compositional Rollout Mismatch in Neuro-Symbolic Search
TAME the BALROG: Task-Adaptive Modular Evolution framework for Game Agents
The Compositional Generalization Gap in Named Entity Recognition: Static Benchmarks Overestimate Transferable Performance
THEIA: Learning Complete Kleene Three-Valued Logic in a Pure-Neural Modular Architecture
The Theory and Practice of MAP Inference over Non-Convex Constraints
Toward Compositional Latent Action Interfaces for Generalizable Agents
Tracking Training Phases in Compositional Learning with Task-Agnostic Measures
Unsafe Only in Combination: Interaction-Barrier Shielding for Tool-Using LLM Agents
Unsupervised Decomposition with Recombination-Consistent Diffusion Models
VASAE: Naming SAE Dictionary Directions with Vocabulary-Aligned Anchoring
Visual Counterfactual Explanations with Compositional Generative Models
What Do Latent Agents Actually Represent? Interpreting Hidden-State Communication in Multi-Agent Systems
What makes the whole? Probing Attribute-Level Compositionality in LLM Judges
When Do Diffusion Models learn to Generate Multiple Objects?
When Does Disentanglement Enable Compositional Generalization? A Transfer Bound and Its Empirical Validation
When Do Multi-Agent Systems Outperform? Analysing the Learning Efficiency of Agentic Systems
When GUI Grounding Fails: Entropy-Based Analysis and Training-Free Refinement
When to Re-Plan: Subgoal Persistence in Hierarchical Latent Reasoning
Where’s the Plan? Locating Latent Planning in Language Models with Lightweight Mechanistic Interventions
Which Way Did It Move? Diagnosing and Overcoming Directional Motion Blindness in Video LLMs
Your Agent, Their Asset: A Real-World Safety Analysis of OpenClaw

« prev Page 3/3