SkillCAT: New Framework Enables Language Model Agents to Develop Reusable Skills Without Training
Researchers have developed SkillCAT, a training-free framework that helps large language model agents learn and refine reusable skills from task execution trajectories. The method uses contrastive analysis to identify what causes success or failure, tests skill improvements before merging them, and loads only relevant skills during inference. The approach achieved up to 40.40% improvement over existing baselines across multiple benchmarks including spreadsheet tasks, table question-answering, and document analysis.
SkillCAT addresses limitations in current skill self-evolution methods for LLM agents by introducing a three-stage process that improves how agents learn from experience. The framework uses Contrastive Causal Extraction to analyze multiple task trajectories and compare successful versus failed attempts to identify key differences, Assessment-Augmented Evolution to validate skill improvements before merging them, and Topology-Aware Task Execution to organize skills into a routable structure that loads only task-relevant capabilities during inference. The researchers evaluated SkillCAT on established benchmarks including SpreadsheetBench, WikiTableQuestions, and DocVQA, and tested its ability to generalize across different models and out-of-distribution scenarios. Results demonstrated consistent improvements, with average scores rising up to 40.40% compared to baseline methods, while maintaining the advantage of requiring no model retraining.
What's missing
The paper does not discuss computational costs or inference time comparisons with baseline methods, nor does it address potential limitations of the approach such as failure modes or scenarios where the framework may underperform.
What different sources said
- arXiv cs.CLCenter
SkillCAT: Contrastive Assessment and Topology-Aware Skill Self-Evolution for LLM Agents
Related
Topology-Aware Thermodynamics Improves DNA Probe Specificity Design
Researchers developed a new framework for designing DNA probes that accounts for the spatial organization of matched sequences, not just overall thermodynamic stability. Traditional methods rely on scalar measures like melting temperature and free energy, which miss how mismatches are distributed along the probe. The approach could improve diagnostic accuracy in applications like HPV detection and gene expression profiling.
Study Identifies Optimal Thermal Dose for Combining Focused Ultrasound with Immunotherapy in Tumors
Researchers used multimodal PET imaging to identify an optimal thermal dose range for focused ultrasound ablation that destroys tumor tissue while preserving conditions for immunotherapy delivery. The study found that excessive heating collapses blood vessels needed for antibody access, while insufficient heating fails to adequately reduce tumor burden. The findings could guide clinical design of combination treatments pairing thermal ablation with immunotherapies.
Plant MSH1 Protein Functions as Mismatch-Directed Nuclease for Organelle Genome Maintenance
Researchers have identified the precise mechanism by which the AtMSH1 protein in Arabidopsis plants recognizes and cleaves DNA mismatches and lesions, preventing mutations in organellar genomes. The protein combines a DNA mismatch recognition module with a nuclease domain that makes staggered cuts at specific positions relative to DNA damage. This discovery explains how plants maintain unusually low mutation rates in their mitochondrial and chloroplast DNA compared to other eukaryotes.