A Unified Per-Token Gating Family for On-Policy Distillation: FKL/RKL Mixing with Multi-Channel and Bias Coefficients
1mo
- Published
- Collected
Authors
Suwan Wu, Yumeng Lin, Pengcheng Yuan, Xiaolong Jiang
Categories
Per-token gating of forward/reverse KL losses has become a standard technique for on-policy knowledge distillation (OPD), but existing methods such as EOPD (Jin et al., 2026) and ToDi (Jung et al., 2025) each fix a single gating signal and a single gating direction, and the two have never been compared directly. We introduce a four-coefficient parameterization lambda_t = sigma(a * h_t + b * u(x) + c + d * gap_t) in which direction-aligned proxies of EOPD and ToDi appear as one-dimensional (1D) restrictions, and which adds multi-channel composition and an explicit bias as further degrees of freedom. On TweetEval (Barbieri et al., 2020) emotion and hate, with a Qwen3-32B teacher and a Qwen3-4B student, configurations in the full family reach higher accuracy than the matched-magnitude single-channel (entropy-only / gap-only) 1D restrictions in 33 of 36 comparable cells, and a 26-cell mean-match isolation experiment places dynamic gating ahead of effective-KL-matched static baselines in 19 of 26 cells. Because cells share training data, models, and parameter substructure, we report both counts as exploratory aggregate directional evidence rather than as independent hypothesis tests. Targeted three-seed paired replications of the nine headline comparisons singled out by that sweep -- including a third task, offensive -- are directionally consistent, but individually smaller than the single-seed estimates and not significant at n=3. We therefore present the parameterization primarily as a shared coordinate system for comparing per-token gating designs in short-output classification OPD.
overfeed.news indexes and links. We publish a short excerpt — the full article stays at arXiv AI Papers.
More from arXiv AI Papers
Log in to follow this source
Generative AI in Turbulence: Foundations, Applications, Validations and Benchmarks
Generative AI offers new approaches to turbulence problems in which incomplete observations or unresolved d.o.f. admit multiple physically plausible solutions. Potential applications include stochastic closures, super-resolution, data assimilation, uncertainty quantification, and the generation of Eulerian and Lagrangian flow ensembles. However, turbulence is not a conventional image-generation problem. Its multiscale dynamics, intermittency, extreme events, rough velocity fields and intrinsically stochastic transport make visual realism, pointwise error, and low-order spectral agreement insufficient measures of physical fidelity. We examine where generative AI provide novel capabilities with emphasis on diffusion models, score-based methods, and related probability-transport formulations. We discuss their use in unconditional generation, probabilistic reconstruction, reduced-order modeling, temporal evolution, and mesh-free representations. We argue that validation must address turbulence properties: energy transfer, structure functions, anomalous scaling, pdf tails, coherent structures, conservation laws, temporal consistency, ensemble diversity and generalization. Finally, we discuss potential limitations in validation, baseline, data availability and reproducibility, and the need for a benchmark ecosystem combining open data, standardized challenges, hidden tests, physics-aware metrics, and documented failures. This paper serves as a bridge between two communities. For AI practitioners, we provide a fair and accessible introduction to turbulence, essential for appreciating the challenges posed by the field and for assessing the physical relevance of generative models. For turbulent practitioners, we provide an introduction to the key technological aspects of Generative AI, clarifying what these tools actually do, and where their potential and current limitations lie.
Quality over Quantity: Rethinking Data and Model Assumptions for Bengali Coreference Resolution
Multilingual transformer models are often assumed to generalise across languages, but their effectiveness for coreference resolution in low-resource settings remains unverified. We investigate this gap in Bengali, a language with over 250 million speakers but limited annotated coreference resources. We evaluate transformer-based pre-trained encoders under multiple settings, including zero-shot, fine-tuning, data augmentation, and cross-lingual transfer. Our results challenge several common assumptions about model generalisation and data efficiency in low-resource NLP. We demonstrate that a small amount of high-quality annotated data significantly outperforms larger mixed-quality datasets (p < 0.01). We find that data augmentation, particularly back-translation, does not improve performance and can degrade results in coreference resolution. Cross-lingual transfer is also inconsistent: transfer from a distant language like English reduces performance, and transfer from the related language Hindi provides no consistent benefit across models. Several widely used multilingual models exhibit collapsed embedding spaces, rendering them unsuitable for similarity-based clustering tasks. These findings highlight the importance of annotation quality over data quantity and provide empirical guidance for developing coreference resolution systems in low-resource languages.
UTA: Universal Training Adaptive Learning Rate Decay
Learning rate decay (LRD) is critical for training high-performance models, yet most designs remain empirically driven. Although the Universal Budget Aware (UBA) LRD offers a theoretical framework, its reliance on SGD dynamics limits its applicability to modern Adam-type optimizers and leaves its key `optimization difficulty parameter' opaque. To bridge this gap, we propose the Universal Training Adaptive (UTA) LRD. Unlike UBA's fixed offline trajectory anchored solely to iterations, UTA treats LRD as a closed-loop shape-and-scale problem. By continuous-time formulation, UTA is training-state adaptive with segment-wise optimal shape. This adaptivity links external iterations with internal training states, enabling a near-optimal path in both progression shape and learning rate (LR) scale, a property unattainable by other LRDs, including UBA. Empirically, we demonstrate consistent improvements across language models (LLaMA 60M to 1B) and vision models (ResNet50, ViT-B) on ImageNet with plug-and-play ease and negligible overhead. Extensive ablation studies further confirm UTA's robustness across dimensions, including peak LR, final decay ratio, its parameter _0, component efficacy and generalization to fine-tuning scenarios, underscoring its practical implementability.
Subspace Crystallization in Grokking: Rank Compression Tracks the Phase Transition
Grokking describes understanding something deeply and intuitively. In neural networks, it is the phenomenon where a network generalizes long after memorizing its training data. This delay has been linked to a fall in the effective rank of the network's weight matrices, but whether this rank compression drives grokking or follows it is still unresolved. We study a mean-field MLP and a one-layer transformer trained on modular arithmetic and observe that: (i) Compression follows the onset, not memorization. Every transformer memorizes its training data within 250 steps, while the grokking onset (the first step at which test accuracy reaches 95%) varies from 6,500 to 21,000 steps. The steepest fall in effective rank comes about 900 steps after the onset, however late the onset occurs, and networks that memorize without grokking compress far less. (ii) Subspace crystallization. We call the span of the fewest top singular directions of a weight matrix that preserve 99% of test accuracy its sufficient subspace. Its size can shrink as the margin between the correct and the best wrong output grows, so we also count the directions needed to keep the outputs within 10%. In the transformer, this count falls during the transition from a quarter or more of the available directions to about a tenth, which we call subspace crystallization. (iii) Imposing low rank does not speed up grokking. Against a factorized full-rank control, capping the rank of the transformer's MLP input weights delays the onset, and capping the rank of its embedding stops 15 of 20 runs from grokking. Compression is therefore a signature of grokking rather than its cause.