The paper studies weak-to-strong generalization (W2SG), where a student model trained on a weaker teacher’s labels surpasses the teacher on the target task. Using a Bregman divergence bias‑variance decomposition, it shows that the student‑teacher risk gap depends on their expected misfit, without requiring convexity of the student hypothesis class. For squared loss, a sufficient condition is that the student converges to the teacher’s posterior mean, achievable by enlarging the student; for cross‑entropy loss, reducing the student’s predictive entropy and using reverse cross‑entropy can promote W2SG, which is empirically validated.
arXiv:2503. 08038v2 Announce Type: replace-cross Abstract: In this paper, we delve deeper into the Kullback-Leibler (KL) Divergence loss and mathematically prove that it is equivalent to the Decoupled Kullback-Leibler (DKL) Divergence loss that consists of (1) a weighted Mean Square Error (wMSE) loss and (2) a Cross-Entropy loss incorporating soft labels.
arXiv:2609.36734v1 Announce Type: new Abstract: Knowledge Distillation (KD) trains a smaller-capacity student model to imitate a larger-capacity teacher model by matching output distributions, implic...
arXiv:2501.18581v5 Announce Type: replace Abstract: Bias-variance decompositions are widely used to understand the generalization performance of machine learning models. While the squared error loss...
arXiv:2607. 15450v1 Announce Type: cross Abstract: Self-distillation (SD) is typically studied when the student is retrained on the teacher's original training inputs.
The paper investigates how to choose the best quantized model from a family of compressed versions when target labels are scarce or unavailable. It finds that a simple rule based on minimum teacher distortion consistently selects the same eight‑bit, per‑channel, unclipped configuration, though this does not minimize empirical target cross‑entropy. The study also shows that confidence‑based estimators perform poorly in overconfident regimes, while output‑distribution estimators can outperform the teacher in some architectures, and that combining distortion with a supervised term can improve selection. Across 134 candidate families, teacher‑anchored selection reduces mean regret with very few labels, though the benefit diminishes after about 25 labels.
arXiv:2602. 04402v3 Announce Type: replace-cross Abstract: Performative predictions influence the very outcomes they aim to forecast.
arXiv:2609.22566v1 Announce Type: cross Abstract: Knowledge distillation (KD) aims to compress high-performance teacher LLMs into lightweight students. However, distilled students often exhibit subst...
arXiv:2601. 01484v2 Announce Type: replace Abstract: Knowledge Distillation (KD) is a central paradigm for transferring knowledge from a large teacher network to a typically smaller student model, often by leveraging soft probabilistic outputs.
arXiv:2607. 02502v1 Announce Type: cross Abstract: On-policy self-distillation (OPSD) has emerged as a practical method for training large language models (LLMs) to reason, where a single model acts as both the teacher and the student with different levels of information access.
The paper investigates how the sampled-token reverse-KL loss in on‑policy distillation distributes updates across tokens. By analyzing the gradient of the per‑token K2 estimator, the authors find that tokens with low student probability and large teacher‑student gaps receive disproportionately large gradient norms. They propose Surprise‑aware Reweighting (SuRe), a lightweight weighting rule that further amplifies this allocation, and demonstrate that SuRe improves math metrics on Qwen3 student models without harming out‑of‑domain performance.
arXiv:2609.31101v1 Announce Type: cross Abstract: The flatness of the loss landscape at a minimizer is a widely used heuristic for reasoning about neural-network generalization, yet evidence for this...
arXiv:2607. 26246v1 Announce Type: new Abstract: On-policy distillation (OPD), which aligns a student with the teacher's token-level distribution on the student's own rollouts, is an effective paradigm for transferring capabilities across LLMs.