paper-with-me

홈 › Papers

SmolTulu: Higher Learning Rate to Batch Size Ratios Can Lead to Better Reasoning in SLMs

2024-12-11 · Sultan Alrashed

We present SmolTulu-1.7b-Instruct, referenced in this report as SmolTulu-DPO-1130, an instruction-tuned language model that adapts AllenAI's Tulu 3 post-training pipeline to enhance Huggingface's SmolLM2-1.7B base model. Through comprehensive empirical analysis using a 135M parameter model, we demonstrate that the relationship between learning rate and batch size significantly impacts model performance in a task-dependent manner. Our findings reveal a clear split: reasoning tasks like ARC and GSM8K benefit from higher learning rate to batch size ratios, while pattern recognition tasks such as HellaSwag and IFEval show optimal performance with lower ratios. These insights informed the development of SmolTulu, which achieves state-of-the-art performance among sub-2B parameter models on instruction following, scoring 67.7% on IFEval ($\Delta$11%), and mathematical reasoning with 51.6% on GSM8K ($\Delta$3.4%), with an alternate version achieving scoring 57.1% on ARC ($\Delta5.4%$). We release our model, training recipes, and ablation studies to facilitate further research in efficient model alignment, demonstrating that careful adaptation of optimization dynamics can help bridge the capability gap between small and large language models.

📄 PDF Abstract BibTeX arXiv:2412.08347

Code (0)

등록된 구현이 없습니다.

Tasks

ARCGSM8KHellaSwagInstruction FollowingLanguage ModelingLanguage ModellingMathematical Reasoning

Methods 이 논문이 사용한 방법론

BASE 설명 없음

Similar Papers 제목 키워드 기반

Scaling Off-Policy Reinforcement Learning with Batch and Weight Normalization

2025-02-11 · Daniel Palenicek, Florian Vogt, Jan Peters

Reinforcement learning has achieved significant milestones, but sample efficiency remains a bottleneck for real-world applications. Recently, CrossQ has demonstrated state-of-the-art sample efficiency with a low update-t…

continuous-controlContinuous Controlreinforcement-learningReinforcement Learning

Improving Layer-wise Adaptive Rate Methods using Trust Ratio Clipping

2020-11-27 · Jeffrey Fong, Siwei Chen, Kaiqi Chen

Training neural networks with large batch is of fundamental significance to deep learning. Large batch training remarkably reduces the amount of training time but has difficulties in maintaining accuracy. Recent works ha…

image-classificationImage Classification

MERIT: Maximum-normalized Element-wise Ratio for Language Model Large-batch Training

2025-08-28 · Yang Luo, Zangwei Zheng, Ziheng Qin, Zirui Zhu 외 arxiv

Large-batch training has become a cornerstone in accelerating the training of deep neural networks, yet it poses challenges in optimization and generalization. Existing optimizers like AdamW present performance degradati…

The Design Space of Tri-Modal Masked Diffusion Models

2026-02-25 · Louis Bethune, Victor Turrisi, Bruno Kacper Mlodozeniec, Pau Rodriguez Lopez 외 arxiv

Discrete diffusion models have emerged as strong alternatives to autoregressive language models, with recent work initializing and fine-tuning a base unimodal model for bimodal generation. Diverging from previous approac…

Stochastic OptimizationText Generation

Nemotron-Flash: Towards Latency-Optimal Hybrid Small Language Models

2025-11-24 · Yonggan Fu, Xin Dong, Shizhe Diao, Matthijs Van keirsbilck 외 arxiv

Efficient deployment of small language models (SLMs) is essential for numerous real-world applications with stringent latency constraints. While previous work on SLM design has primarily focused on reducing the number of…