Enhancing Cross-Tokenizer Knowledge Distillation with Contextual Dynamical Mapping

• URL: http://arxiv.org/abs/2502.11104v1
• Date: Sun, 16 Feb 2025 12:46:07 GMT
• Title: Enhancing Cross-Tokenizer Knowledge Distillation with Contextual Dynamical Mapping
• Authors: Yijie Chen, Yijin Liu, Fandong Meng, Yufeng Chen, Jinan Xu, Jie Zhou,
• Abstract summary: Contextual Dynamic Mapping (CDM) is a novel cross-tokenizer distillation framework.<n>It uses contextual information to enhance sequence alignment precision and dynamically improve vocabulary mapping.<n>Our method shows significant advantages over existing cross-tokenizer distillation baselines across diverse benchmarks.
• Score: 85.48043537327258
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Knowledge Distillation (KD) has emerged as a prominent technique for model compression. However, conventional KD approaches primarily focus on homogeneous architectures with identical tokenizers, constraining their applicability in cross-architecture scenarios. As for the cross-tokenizer KD, the differences in the tokenizers give rise to two fundamental challenges: (1) sequence misalignment caused by divergent tokenization strategies, and (2) mismatched vocabulary size and composition. While existing probability-matching methods attempt to address these issues, their efficacy remains limited due to suboptimal alignment in both the sequence and vocabulary aspects. To overcome these limitations, we propose Contextual Dynamic Mapping (CDM), a novel cross-tokenizer distillation framework that employs contextual information to enhance sequence alignment precision and dynamically improves vocabulary mapping. We evaluated the effectiveness of our approach across five advanced and widely-used model families (i.e, LLama3, Phi3, Gemma2, OPT and Qwen2), which were configured into three distinct teacher-student pairs. Our method shows significant advantages over existing cross-tokenizer distillation baselines across diverse benchmarks, including instruction-following, code generation and math. Notably, our analysis reveals that combining conventional same-tokenizer distillation and cross-tokenizer distillation through CDM yields further performance improvements. The code is available at https://github.com/pppa2019/ContexualDynamicMapping

Related papers

• Cross-Tokenizer Distillation via Approximate Likelihood Matching [17.597293085255075]
  We develop a cross-tokenizer distillation method to solve this deficiency. Our method is the first to enable cross-tokenizer distillation without a next-token prediction loss. Our results make substantial strides toward better adaptability and enhanced interaction between different LLMs.
  arXiv  Detail & Related papers  (2025-03-25T21:44:10Z)
• CoT2Align: Cross-Chain of Thought Distillation via Optimal Transport Alignment for Language Models with Different Tokenizers [45.59157559718677]
  Large Language Models (LLMs) achieve state-of-the-art performance across various NLP tasks but face deployment challenges due to high computational costs and memory constraints. Knowledge distillation (KD) is a promising solution, transferring knowledge from large teacher models to smaller student models. We propose CoT2Align, a universal KD framework that integrates Chain-of-Thought (CoT) augmentation and introduces Cross-CoT Alignment to enhance reasoning transfer.
  arXiv  Detail & Related papers  (2025-02-24T03:30:29Z)
• Multi-Level Optimal Transport for Universal Cross-Tokenizer Knowledge Distillation on Language Models [81.74999702045339]
  Multi-Level Optimal Transport (MultiLevelOT) is a novel approach that advances the optimal transport for universal cross-tokenizer knowledge distillation.<n>Our method aligns the logit distributions of the teacher and the student at both token and sequence levels.<n>At the token level, MultiLevelOT integrates both global and local information by jointly optimizing all tokens within a sequence to enhance robustness.
  arXiv  Detail & Related papers  (2024-12-19T04:51:06Z)
• BEST-STD: Bidirectional Mamba-Enhanced Speech Tokenization for Spoken Term Detection [8.303512060791736]
  Spoken term detection is often hindered by reliance on frame-level features and the computationally intensive DTW-based template matching.<n>We propose a novel approach that encodes speech into discrete, speaker-agnostic semantic tokens.<n>This facilitates fast retrieval using text-based search algorithms and effectively handles out-of-vocabulary terms.
  arXiv  Detail & Related papers  (2024-11-21T13:05:18Z)
• SGC-VQGAN: Towards Complex Scene Representation via Semantic Guided Clustering Codebook [9.993066868670283]
  We introduce SGC-VQGAN through Semantic Online Clustering method to enhance token semantics through Consistent Semantic Learning. Our approach constructs a temporospatially consistent semantic codebook, addressing issues of codebook collapse and imbalanced token semantics.
  arXiv  Detail & Related papers  (2024-09-09T23:12:43Z)
• Bit Cipher -- A Simple yet Powerful Word Representation System that Integrates Efficiently with Language Models [4.807347156077897]
  Bit-cipher is a word representation system that eliminates the need of backpropagation and hyper-efficient dimensionality reduction techniques. We perform probing experiments on part-of-speech (POS) tagging and named entity recognition (NER) to assess bit-cipher's competitiveness with classic embeddings. By replacing embedding layers with cipher embeddings, our experiments illustrate the notable efficiency of cipher in accelerating the training process and attaining better optima.
  arXiv  Detail & Related papers  (2023-11-18T08:47:35Z)
• Quick Dense Retrievers Consume KALE: Post Training Kullback Leibler Alignment of Embeddings for Asymmetrical dual encoders [89.29256833403169]
  We introduce Kullback Leibler Alignment of Embeddings (KALE), an efficient and accurate method for increasing the inference efficiency of dense retrieval methods. KALE extends traditional Knowledge Distillation after bi-encoder training, allowing for effective query encoder compression without full retraining or index generation. Using KALE and asymmetric training, we can generate models which exceed the performance of DistilBERT despite having 3x faster inference.
  arXiv  Detail & Related papers  (2023-03-31T15:44:13Z)
• Keyphrase Extraction with Dynamic Graph Convolutional Networks and Diversified Inference [50.768682650658384]
  Keyphrase extraction (KE) aims to summarize a set of phrases that accurately express a concept or a topic covered in a given document. Recent Sequence-to-Sequence (Seq2Seq) based generative framework is widely used in KE task, and it has obtained competitive performance on various benchmarks. In this paper, we propose to adopt the Dynamic Graph Convolutional Networks (DGCN) to solve the above two problems simultaneously.
  arXiv  Detail & Related papers  (2020-10-24T08:11:23Z)
• Boosting Continuous Sign Language Recognition via Cross Modality Augmentation [135.30357113518127]
  Continuous sign language recognition deals with unaligned video-text pair. We propose a novel architecture with cross modality augmentation. The proposed framework can be easily extended to other existing CTC based continuous SLR architectures.
  arXiv  Detail & Related papers  (2020-10-11T15:07:50Z)

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.