Uncovering Intrinsic Capabilities: A Paradigm for Data Curation in Vision-Language Models
Large vision-language models (VLMs) achieve strong benchmark performance, but controlling their behavior through instruction tuning remains difficult. Reducing the budget of instruction tuning dataset often causes regressions, as heuristic strategies treat models as black boxes and overlook the latent capabilities that govern learning. We introduce Capability-Attributed Data Curation (CADC), a framework that shifts curation from task-specific heuristics to intrinsic capability analysis. CADC discovers intrinsic capabilities in an unsupervised manner from gradient-based learning trajectories, attributes training data to these capabilities via influence estimation, and curates capability-aware curricula through balanced selection and staged sequencing. This transforms black-box instruction tuning into a controllable, capability-driven process. With as little as 5% of the original data, CADC surpasses full-data training on multimodal benchmarks. These results validate intrinsic capabilities as the fundamental building blocks of model learning and establish CADC as a principle paradigm for instruction data curation.
Code (0)
등록된 구현이 없습니다.
Similar Papers 제목 키워드 기반
AnE: Pushing the Reasoning Frontier of Multimodal LLMs via Anchor Evolution
Post-training via Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) is crucial for enhancing reasoning in Multimodal Large Language Models (MLLMs), yet existing paradigms often reach a performance bottleneck d…
Reinforcement LearningMultimodal ReasoningTracing the Roots: A Multi-Agent Framework for Uncovering Data Lineage in Post-Training LLMs
Post-training data plays a pivotal role in shaping the capabilities of Large Language Models (LLMs), yet datasets are often treated as isolated artifacts, overlooking the systemic connections that underlie their evolutio…
A Good CREPE needs more than just Sugar: Investigating Biases in Compositional Vision-Language Benchmarks
We investigate 17 benchmarks (e.g. SugarCREPE, VALSE) commonly used for measuring compositional understanding capabilities of vision-language models (VLMs). We scrutinize design choices in their construction, including d…
Language ModelingLanguage ModellingBest practices for the manual curation of Intrinsically Disordered Proteins in DisProt
The DisProt database is a significant resource containing manually curated data on experimentally validated intrinsically disordered proteins (IDPs) and regions (IDRs) from the literature. Developed in 2005, its primary …
AI4D - African Language Dataset Challenge
As language and speech technologies become more advanced, the lack of fundamental digital resources for African languages, such as data, spell checkers and PoS taggers, means that the digital divide between these languag…
POS