Recursive Inference Scaling: A Winning Path to Scalable Inference in Language and Multimodal Systems

• URL: http://arxiv.org/abs/2502.07503v2
• Date: Wed, 19 Feb 2025 09:24:45 GMT
• Title: Recursive Inference Scaling: A Winning Path to Scalable Inference in Language and Multimodal Systems
• Authors: Ibrahim Alabdulmohsin, Xiaohua Zhai,
• Abstract summary: We introduce Recursive INference Scaling (RINS) as a complementary, plug-in recipe for scaling inference time.<n>For a given fixed model architecture and training compute budget, RINS substantially improves language modeling performance.<n>RINS delivers gains in multimodal systems, including a +2% improvement in 0-shot ImageNet accuracy for SigLIP-B/16.
• Score: 21.01887711305712
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Recent research in language modeling reveals two scaling effects: the well-known improvement from increased training compute, and a lesser-known boost from applying more sophisticated or computationally intensive inference methods. Inspired by recent findings on the fractal geometry of language, we introduce Recursive INference Scaling (RINS) as a complementary, plug-in recipe for scaling inference time. For a given fixed model architecture and training compute budget, RINS substantially improves language modeling performance. It also generalizes beyond pure language tasks, delivering gains in multimodal systems, including a +2% improvement in 0-shot ImageNet accuracy for SigLIP-B/16. Additionally, by deriving data scaling laws, we show that RINS improves both the asymptotic performance limits and the scaling exponents. These advantages are maintained even when compared to state-of-the-art recursive techniques like the "repeat-all-over" (RAO) strategy in Mobile LLM. Finally, stochastic RINS not only can enhance performance further but also provides the flexibility to optionally forgo increased inference computation at test time with minimal performance degradation.

Related papers

• Exploring Training and Inference Scaling Laws in Generative Retrieval [50.82554729023865]
  We investigate how model size, training data scale, and inference-time compute jointly influence generative retrieval performance. Our experiments show that n-gram-based methods demonstrate strong alignment with both training and inference scaling laws. We find that LLaMA models consistently outperform T5 models, suggesting a particular advantage for larger decoder-only models in generative retrieval.
  arXiv  Detail & Related papers  (2025-03-24T17:59:03Z)
• C2D-ISR: Optimizing Attention-based Image Super-resolution from Continuous to Discrete Scales [6.700548615812325]
  We propose a novel framework, textbfC2D-ISR, for optimizing attention-based image super-resolution models. Our approach is based on a two-stage training methodology and a hierarchical encoding mechanism. In addition, we generalize the hierarchical encoding mechanism with existing attention-based network structures.
  arXiv  Detail & Related papers  (2025-03-17T21:52:18Z)
• LESA: Learnable LLM Layer Scaling-Up [57.0510934286449]
  Training Large Language Models (LLMs) from scratch requires immense computational resources, making it prohibitively expensive. Model scaling-up offers a promising solution by leveraging the parameters of smaller models to create larger ones. We propose textbfLESA, a novel learnable method for depth scaling-up.
  arXiv  Detail & Related papers  (2025-02-19T14:58:48Z)
• Numerical Pruning for Efficient Autoregressive Models [87.56342118369123]
  This paper focuses on compressing decoder-only transformer-based autoregressive models through structural weight pruning.<n>Specifically, we propose a training-free pruning method that calculates a numerical score with Newton's method for the Attention and modules, respectively.<n>To verify the effectiveness of our method, we provide both theoretical support and extensive experiments.
  arXiv  Detail & Related papers  (2024-12-17T01:09:23Z)
• Inference Scaling for Long-Context Retrieval Augmented Generation [37.15479223789199]
  In this work, we investigate inference scaling for retrieval augmented generation (RAG) We focus on two inference scaling strategies: in-context learning and iterative prompting. We demonstrate that scaling inference compute on long-context large language models achieves up to 58.9% gains on benchmark datasets.
  arXiv  Detail & Related papers  (2024-10-06T03:42:15Z)
• Efficient and Flexible Neural Network Training through Layer-wise Feedback Propagation [49.44309457870649]
  We present Layer-wise Feedback Propagation (LFP), a novel training principle for neural network-like predictors.<n>LFP decomposes a reward to individual neurons based on their respective contributions to solving a given task.<n>Our method then implements a greedy approach reinforcing helpful parts of the network and weakening harmful ones.
  arXiv  Detail & Related papers  (2023-08-23T10:48:28Z)
• Improved Algorithms for Neural Active Learning [74.89097665112621]
  We improve the theoretical and empirical performance of neural-network(NN)-based active learning algorithms for the non-parametric streaming setting. We introduce two regret metrics by minimizing the population loss that are more suitable in active learning than the one used in state-of-the-art (SOTA) related work.
  arXiv  Detail & Related papers  (2022-10-02T05:03:38Z)
• Revisiting Neural Scaling Laws in Language and Vision [43.57394336742374]
  We argue for a more rigorous methodology based on the extrapolation loss, instead of reporting the best-fitting parameters. We present a recipe for estimating scaling law parameters reliably from learning curves. We demonstrate that it extrapolates more accurately than previous methods in a wide range of architecture families across several domains.
  arXiv  Detail & Related papers  (2022-09-13T09:41:51Z)
• Rényi Divergence Deep Mutual Learning [3.682680183777648]
  This paper revisits Deep Learning Mutual (DML) as a simple yet effective computing paradigm. We propose using R'enyi divergence instead of the KL divergence, which is more flexible and limited. Our empirical results demonstrate the advantage combining DML and R'enyi divergence, leading to further improvement in model generalization.
  arXiv  Detail & Related papers  (2022-09-13T04:58:35Z)
• Stabilizing Q-learning with Linear Architectures for Provably Efficient Learning [53.17258888552998]
  This work proposes an exploration variant of the basic $Q$-learning protocol with linear function approximation. We show that the performance of the algorithm degrades very gracefully under a novel and more permissive notion of approximation error.
  arXiv  Detail & Related papers  (2022-06-01T23:26:51Z)
• Dual Optimization for Kolmogorov Model Learning Using Enhanced Gradient Descent [8.714458129632158]
  Kolmogorov model (KM) is an interpretable and predictable representation approach to learning the underlying probabilistic structure of a set of random variables. We propose a computationally scalable KM learning algorithm, based on the regularized dual optimization combined with enhanced gradient descent (GD) method. It is shown that the accuracy of logical relation mining for interpretability by using the proposed KM learning algorithm exceeds $80%$.
  arXiv  Detail & Related papers  (2021-07-11T10:33:02Z)
• Extrapolation for Large-batch Training in Deep Learning [72.61259487233214]
  We show that a host of variations can be covered in a unified framework that we propose. We prove the convergence of this novel scheme and rigorously evaluate its empirical performance on ResNet, LSTM, and Transformer.
  arXiv  Detail & Related papers  (2020-06-10T08:22:41Z)
• Communication-Efficient Distributed Stochastic AUC Maximization with Deep Neural Networks [50.42141893913188]
  We study a distributed variable for large-scale AUC for a neural network as with a deep neural network. Our model requires a much less number of communication rounds and still a number of communication rounds in theory. Our experiments on several datasets show the effectiveness of our theory and also confirm our theory.
  arXiv  Detail & Related papers  (2020-05-05T18:08:23Z)
• Generalized Reinforcement Meta Learning for Few-Shot Optimization [3.7675996866306845]
  We present a generic and flexible Reinforcement Learning (RL) based meta-learning framework for the problem of few-shot learning. Our framework could be easily extended to do network architecture search.
  arXiv  Detail & Related papers  (2020-05-04T03:21:05Z)

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.