Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Policy Compatible Skill Incremental Learning via Lazy Learning Interface
Authors: Daehee Lee, Dongsu Lee, TaeYoon Kwack, Wonje Choi, Honguk Woo
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate SIL-C across diverse SIL scenarios and demonstrate that it maintains compatibility between evolving skills and downstream policies while ensuring efficiency throughout the learning process. Source code: https://github.com/L2dulgi/SIL-C |
| Researcher Affiliation | Academia | Sungkyunkwan University University of Texas at Austin EMAIL, EMAIL |
| Pseudocode | Yes | A.3 Pseudocode |
| Open Source Code | Yes | Source code: https://github.com/L2dulgi/SIL-C |
| Open Datasets | Yes | To evaluate skill-policy compatibility, we construct various SIL scenarios using two simulation environments: Franka Kitchen [65, 66] and Meta-World [67, 11]. |
| Dataset Splits | Yes | In each phase p, a new skill dataset Dp is provided to train the skill decoder. Subsequently, based on the trained low-level skill decoder, 24 task-specific high-level policies are individually trained. ... Experience Replay (ER) [69]: In our implementation, a replay buffer stores 10% of the data from each previous phase. During training in the next phase, these stored samples are interleaved with new data at a 1:1 sampling ratio to mitigate forgetting. |
| Hardware Specification | Yes | AMD Ryzen 9 7950X3D 16-Core Processor with a single RTX 4090 GPU. OS: Ubuntu 22.04, CUDA Version: 12.4, Driver Version: 550.144.03 |
| Software Dependencies | Yes | The versions of core libraries are: jax: 0.4.34 jaxlib: 0.4.34 flax: 0.10.2 optax: 0.1.9 |
| Experiment Setup | Yes | Table 5: Default hyperparameter configuration for SIL-C Table 10: Default policy configuration Table 11: Default skill decoder configuration |