On the Predictive Power of Representation Dispersion in Language Models

• URL: http://arxiv.org/abs/2506.24106v1
• Date: Mon, 30 Jun 2025 17:53:50 GMT
• Title: On the Predictive Power of Representation Dispersion in Language Models
• Authors: Yanhong Li, Ming Li, Karen Livescu, Jiawei Zhou,
• Abstract summary: We show that models that spread contextual representations more widely tend to achieve lower perplexity.<n>We show how dispersion can be leveraged for a range of practical tasks without requiring labeled data.
• Score: 27.85164935773766
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: We show that a language model's ability to predict text is tightly linked to the breadth of its embedding space: models that spread their contextual representations more widely tend to achieve lower perplexity. Concretely, we find that representation dispersion - the average pairwise cosine distance among hidden vectors - strongly and negatively correlates with perplexity across diverse model families (LLaMA, Qwen, and others) and domains (Wikipedia, news, scientific abstracts). Beyond illustrating this link, we show how dispersion can be leveraged for a range of practical tasks without requiring labeled data. First, measuring dispersion on unlabeled text allows us to predict downstream accuracy in new domains, offering a data-efficient tool for model selection. Next, we find that identifying layers with higher dispersion pinpoints the best representations for retrieval-based methods such as kNN-LM, bypassing exhaustive layer-by-layer searches. Finally, we integrate a simple push-away objective into training, which increases dispersion in both single-domain and cross-domain scenarios and directly improves perplexity in each.

Related papers

• Theory on Score-Mismatched Diffusion Models and Zero-Shot Conditional Samplers [49.97755400231656]
  We present the first performance guarantee with explicit dimensional dependencies for general score-mismatched diffusion samplers.<n>We show that score mismatches result in an distributional bias between the target and sampling distributions, proportional to the accumulated mismatch between the target and training distributions.<n>This result can be directly applied to zero-shot conditional samplers for any conditional model, irrespective of measurement noise.
  arXiv  Detail & Related papers  (2024-10-17T16:42:12Z)
• Downstream-Pretext Domain Knowledge Traceback for Active Learning [138.02530777915362]
  We propose a downstream-pretext domain knowledge traceback (DOKT) method that traces the data interactions of downstream knowledge and pre-training guidance. DOKT consists of a traceback diversity indicator and a domain-based uncertainty estimator. Experiments conducted on ten datasets show that our model outperforms other state-of-the-art methods.
  arXiv  Detail & Related papers  (2024-07-20T01:34:13Z)
• Diffusion Features to Bridge Domain Gap for Semantic Segmentation [2.8616666231199424]
  This paper investigates the approach that leverages the sampling and fusion techniques to harness the features of diffusion models efficiently. By leveraging the strength of text-to-image generation capability, we introduce a new training framework designed to implicitly learn posterior knowledge from it.
  arXiv  Detail & Related papers  (2024-06-02T15:33:46Z)
• Amortizing intractable inference in diffusion models for vision, language, and control [89.65631572949702]
  This paper studies amortized sampling of the posterior over data, $mathbfxsim prm post(mathbfx)propto p(mathbfx)r(mathbfx)$, in a model that consists of a diffusion generative model prior $p(mathbfx)$ and a black-box constraint or function $r(mathbfx)$.<n>We prove the correctness of a data-free learning objective, relative trajectory balance, for training a diffusion model that samples from
  arXiv  Detail & Related papers  (2024-05-31T16:18:46Z)
• Deep Generative Sampling in the Dual Divergence Space: A Data-efficient & Interpretative Approach for Generative AI [29.13807697733638]
  We build on the remarkable achievements in generative sampling of natural images. We propose an innovative challenge, potentially overly ambitious, which involves generating samples that resemble images. The statistical challenge lies in the small sample size, sometimes consisting of a few hundred subjects.
  arXiv  Detail & Related papers  (2024-04-10T22:35:06Z)
• Distributed Markov Chain Monte Carlo Sampling based on the Alternating Direction Method of Multipliers [143.6249073384419]
  In this paper, we propose a distributed sampling scheme based on the alternating direction method of multipliers. We provide both theoretical guarantees of our algorithm's convergence and experimental evidence of its superiority to the state-of-the-art. In simulation, we deploy our algorithm on linear and logistic regression tasks and illustrate its fast convergence compared to existing gradient-based methods.
  arXiv  Detail & Related papers  (2024-01-29T02:08:40Z)
• Implicit Variational Inference for High-Dimensional Posteriors [7.924706533725115]
  In variational inference, the benefits of Bayesian models rely on accurately capturing the true posterior distribution. We propose using neural samplers that specify implicit distributions, which are well-suited for approximating complex multimodal and correlated posteriors. Our approach introduces novel bounds for approximate inference using implicit distributions by locally linearising the neural sampler.
  arXiv  Detail & Related papers  (2023-10-10T14:06:56Z)
• Prompting Diffusion Representations for Cross-Domain Semantic Segmentation [101.04326113360342]
  diffusion-pretraining achieves extraordinary domain generalization results for semantic segmentation. We introduce a scene prompt and a prompt randomization strategy to help further disentangle the domain-invariant information when training the segmentation head.
  arXiv  Detail & Related papers  (2023-07-05T09:28:25Z)
• Score Approximation, Estimation and Distribution Recovery of Diffusion Models on Low-Dimensional Data [68.62134204367668]
  This paper studies score approximation, estimation, and distribution recovery of diffusion models, when data are supported on an unknown low-dimensional linear subspace. We show that with a properly chosen neural network architecture, the score function can be both accurately approximated and efficiently estimated. The generated distribution based on the estimated score function captures the data geometric structures and converges to a close vicinity of the data distribution.
  arXiv  Detail & Related papers  (2023-02-14T17:02:35Z)
• Learning Invariant Representations and Risks for Semi-supervised Domain Adaptation [109.73983088432364]
  We propose the first method that aims to simultaneously learn invariant representations and risks under the setting of semi-supervised domain adaptation (Semi-DA) We introduce the LIRR algorithm for jointly textbfLearning textbfInvariant textbfRepresentations and textbfRisks.
  arXiv  Detail & Related papers  (2020-10-09T15:42:35Z)
• Calibrated Adversarial Refinement for Stochastic Semantic Segmentation [5.849736173068868]
  We present a strategy for learning a calibrated predictive distribution over semantic maps, where the probability associated with each prediction reflects its ground truth correctness likelihood. We demonstrate the versatility and robustness of the approach by achieving state-of-the-art results on the multigrader LIDC dataset and on a modified Cityscapes dataset with injected ambiguities. We show that the core design can be adapted to other tasks requiring learning a calibrated predictive distribution by experimenting on a toy regression dataset.
  arXiv  Detail & Related papers  (2020-06-23T16:39:59Z)
• PushNet: Efficient and Adaptive Neural Message Passing [1.9121961872220468]
  Message passing neural networks have recently evolved into a state-of-the-art approach to representation learning on graphs. Existing methods perform synchronous message passing along all edges in multiple subsequent rounds. We consider a novel asynchronous message passing approach where information is pushed only along the most relevant edges until convergence.
  arXiv  Detail & Related papers  (2020-03-04T18:15:30Z)
• Focus on Semantic Consistency for Cross-domain Crowd Understanding [34.560447389853614]
  Some domain adaptation algorithms try to liberate it by training models with synthetic data. We found that a mass of estimation errors in the background areas impede the performance of the existing methods. In this paper, we propose a domain adaptation method to eliminate it.
  arXiv  Detail & Related papers  (2020-02-20T08:51:05Z)

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.