Capability Salience Vector: Fine-grained Alignment of Loss and Capabilities for Downstream Task Scaling Law

• URL: http://arxiv.org/abs/2506.13216v1
• Date: Mon, 16 Jun 2025 08:16:03 GMT
• Title: Capability Salience Vector: Fine-grained Alignment of Loss and Capabilities for Downstream Task Scaling Law
• Authors: Qiming Ge, Shuhao Xing, Songyang Gao, Yunhua Zhou, Yicheng Zou, Songyang Zhang, Zhi Chen, Hang Yan, Qi Zhang, Qipeng Guo, Kai Chen,
• Abstract summary: We introduce Capability Salience Vector, which decomposes the overall loss and assigns different importance weights to tokens to assess a specific meta-capability.<n>Experiments on various popular benchmarks demonstrate that our proposed Capability Salience Vector could significantly improve the predictability of language model performance on downstream tasks.
• Score: 49.25050966412749
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Scaling law builds the relationship between training computation and validation loss, enabling researchers to effectively predict the loss trending of models across different levels of computation. However, a gap still remains between validation loss and the model's downstream capabilities, making it untrivial to apply scaling law to direct performance prediction for downstream tasks. The loss typically represents a cumulative penalty for predicted tokens, which are implicitly considered to have equal importance. Nevertheless, our studies have shown evidence that when considering different training data distributions, we cannot directly model the relationship between downstream capability and computation or token loss. To bridge the gap between validation loss and downstream task capabilities, in this work, we introduce Capability Salience Vector, which decomposes the overall loss and assigns different importance weights to tokens to assess a specific meta-capability, aligning the validation loss with downstream task performance in terms of the model's capabilities. Experiments on various popular benchmarks demonstrate that our proposed Capability Salience Vector could significantly improve the predictability of language model performance on downstream tasks.

Related papers

• On Fairness of Task Arithmetic: The Role of Task Vectors [1.236974227340167]
  We study how manipulating task vectors affects fairness metrics, including Demographic Parity and Equalized Odds.<n>Our results offer novel insights into the fairness implications of model editing and establish a foundation for fairness-aware and responsible model editing practices.
  arXiv  Detail & Related papers  (2025-05-30T06:34:01Z)
• SINDER: Repairing the Singular Defects of DINOv2 [61.98878352956125]
  Vision Transformer models trained on large-scale datasets often exhibit artifacts in the patch token they extract. We propose a novel fine-tuning smooth regularization that rectifies structural deficiencies using only a small dataset.
  arXiv  Detail & Related papers  (2024-07-23T20:34:23Z)
• Learning Latent Graph Structures and their Uncertainty [63.95971478893842]
  We show that minimizing point-prediction losses does not guarantee proper learning of latent relational information.<n>We propose a sampling-based method that solves this joint learning task.
  arXiv  Detail & Related papers  (2024-05-30T10:49:22Z)
• Predicting Emergent Abilities with Infinite Resolution Evaluation [85.89911520190711]
  We introduce PassUntil, an evaluation strategy with theoretically infinite resolution, through massive sampling in the decoding phase. We predict the performance of the 2.4B model on code generation with merely 0.05% deviation before training starts. We identify a kind of accelerated emergence whose scaling curve cannot be fitted by standard scaling law function.
  arXiv  Detail & Related papers  (2023-10-05T02:35:00Z)
• Expressive Losses for Verified Robustness via Convex Combinations [67.54357965665676]
  We study the relationship between the over-approximation coefficient and performance profiles across different expressive losses. We show that, while expressivity is essential, better approximations of the worst-case loss are not necessarily linked to superior robustness-accuracy trade-offs.
  arXiv  Detail & Related papers  (2023-05-23T12:20:29Z)
• Modeling Uncertain Feature Representation for Domain Generalization [49.129544670700525]
  We show that our method consistently improves the network generalization ability on multiple vision tasks. Our methods are simple yet effective and can be readily integrated into networks without additional trainable parameters or loss constraints.
  arXiv  Detail & Related papers  (2023-01-16T14:25:02Z)
• On Transfer of Adversarial Robustness from Pretraining to Downstream Tasks [1.8900691517352295]
  We show that the robustness of a linear predictor on downstream tasks can be constrained by the robustness of its underlying representation. Our results offer an initial step towards characterizing the requirements of the representation function for reliable post-adaptation performance.
  arXiv  Detail & Related papers  (2022-08-07T23:00:40Z)
• Uncertainty in Contrastive Learning: On the Predictability of Downstream Performance [7.411571833582691]
  We study whether the uncertainty of such a representation can be quantified for a single datapoint in a meaningful way. We show that this goal can be achieved by directly estimating the distribution of the training data in the embedding space.
  arXiv  Detail & Related papers  (2022-07-19T15:44:59Z)
• Mixing between the Cross Entropy and the Expectation Loss Terms [89.30385901335323]
  Cross entropy loss tends to focus on hard to classify samples during training. We show that adding to the optimization goal the expectation loss helps the network to achieve better accuracy. Our experiments show that the new training protocol improves performance across a diverse set of classification domains.
  arXiv  Detail & Related papers  (2021-09-12T23:14:06Z)
• Accurate and Robust Feature Importance Estimation under Distribution Shifts [49.58991359544005]
  PRoFILE is a novel feature importance estimation method. We show significant improvements over state-of-the-art approaches, both in terms of fidelity and robustness.
  arXiv  Detail & Related papers  (2020-09-30T05:29:01Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.