TY  - RPRT
TI  - Decomposing the Basic Abilities of Large Language Models: Mitigating Cross-Task Interference in Multi-Task Instruct-Tuning
AU  - Bing Wang
AU  - Ximing Li
AU  - Changchun Li
AU  - Jinjin Chi
AU  - Gang Niu
AU  - Masashi Sugiyama
PY  - 2026
UR  - https://arxiv.org/abs/2605.05676
ID  - 2605.05676
ER  -