Three papers accepted to ICML 2026! Multimodal Fact-Level Attribution for Verifiable Reasoning, which extends fact-level attribution to multimodal settings to support verifiable reasoning. Skill-Based Mixture-of-Experts: Adaptive Routing for Heterogeneous Reasoning via Inferred Skills, which routes inputs to specialized experts by inferring the skills required for each reasoning task. Generalized Correctness Models: Learning Calibrated and Cross-Model Correctness Predictors from Historical Patterns, which learns correctness predictors that generalize across models by leveraging historical performance patterns.