Bryan Kian Hsiang Low

COBRA-Skills is introduced, an efficient framework that formulates skill optimization as budgeted sequential optimization over a dynamically evolving candidate space and remains robust to changes in the agent harness and performs effectively when the target model itself is used for skill generation and refinement.

Small proxy models can reveal near-optimal SFT-RL budget allocations for large LLMs, streamlining the post-training process significantly.

Forgetting specific data without sacrificing model performance just got a theoretical boost—HAMU guarantees improvements in unlearning quality based on data hardness.

Forget exponential complexity: Adalina slashes the query complexity for approximating Shapley values with a provably adaptive, linear-time, linear-space algorithm.

Forget black-box clustering: WISE delivers state-of-the-art performance on mixed-type data while providing transparent, feature-level explanations grounded in the same primitives that drive clustering.