Look Both Ways and No Sink: Converting LLMs into Text Encoders without Training (2025.acl-long)
Copied to clipboard
| Challenge: | Existing methods for converting large language models into powerful text encoders require extensive training on large datasets. |
| Approach: | They propose a training-free approach that enables bidirectional attention and suppresses the attention sink phenomenon, resulting in superior performance. |
| Outcome: | The proposed approach enables bidirectional attention and suppresses the attention sink phenomenon, resulting in superior performance. |
Similar Papers
Multilingual Encoder Knows more than You Realize: Shared Weights Pretraining for Extremely Low-Resource Languages (2025.acl-long)
Copied to clipboard
| Challenge: | XLM-R and mBART have advanced multilingualism in NLP, but low-resource languages such as Tibetan, Uyghur, Kazakh, and Mongolian are underserved. |
| Approach: | They propose a framework for adapting multilingual encoders to text generation in extremely low-resource languages by reusing the weights between the encoder and the decoder. |
| Outcome: | The proposed framework performs better on various downstream tasks even when compared with much larger models. |
Attention Sinks in Diffusion Language Models (2026.findings-acl)
Copied to clipboard
Maximo Eduardo Rulli, Simone Petruzzi, Edoardo Michielon, Fabrizio Silvestri, Simone Scardapane, Alessio Devoto
| Challenge: | Masked Diffusion Language Models (DLMs) employ transformer encoders with bidirectional attention, enabling parallel token generation while maintaining competitive performance. |
| Approach: | They conduct an empirical analysis of DLM attention patterns focusing on the attention sinking phenomenon . they find that DLMs also exhibit attention sinks, but with distinct characteristics . |
| Outcome: | The proposed models employ transformer encoders with bidirectional attention, enabling parallel token generation while maintaining competitive performance. |
A Training-Free Length Extrapolation Approach for LLMs: Greedy Attention Logit Interpolation (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing solutions to problem of positional out-of-distribution (O.O.D.) are inefficient, redundant, and lack local positional information. |
| Approach: | They propose a training-free method that greedily reuses pretrained positional intervals and interpolates attention logits to eliminate outliers. |
| Outcome: | The proposed method achieves stable and superior performance across long-context tasks without requiring input-length-specific tuning. |
Large Language Models as Foundations for Next-Gen Dense Retrieval: A Comprehensive Empirical Assessment (2024.emnlp-main)
Copied to clipboard
| Challenge: | Pre-trained language models have limited generalization capabilities and performance challenges. |
| Approach: | They evaluate 15 different backbone LLMs and non-LLMs to evaluate their performance . larger models and extensive pre-training consistently enhance in-domain accuracy and data efficiency . |
| Outcome: | The results show that larger models and extensive pre-training enhance in-domain accuracy and data efficiency. |
Data and Model Centric Approaches for Expansion of Large Language Models to New languages (2025.emnlp-tutorials)
Copied to clipboard
| Challenge: | Existing LLMs mainly support English alongside a handful of high resource languages . this leaves a major gap for most low-resource languages despite increasing pace of research . |
| Approach: | This tutorial examines approaches to expand the language coverage of LLMs . they look at tokenizer training, pre-training, instruction tuning, alignment, evaluation, etc. |
| Outcome: | This tutorial examines approaches to expand the language coverage of LLMs . it provides an efficient and viable path to bring LLM technologies to low-resource languages . |
Beyond Decoder-only: Large Language Models Can be Good Encoders for Machine Translation (2025.findings-acl)
Copied to clipboard
Yingfeng Luo, Tong Zheng, Yongyu Mu, Bei Li, Qinghong Zhang, Yongqi Gao, Ziqiang Xu, Peinan Feng, Xiaoqian Liu, Tong Xiao, JingBo Zhu
| Challenge: | Recent advances in machine translation have focused on a single pre-trained decoder . encoder-decoder architectures have received relatively little attention in NMT . |
| Approach: | They propose a method that leverages LLMs as MT encoders and pairs them with lightweight decoders to develop universal translation models. |
| Outcome: | The proposed method matches or surpasses baselines in terms of translation quality but achieves 75% reduction in memory footprint of the KV cache. |
LLMs Are Zero-Shot Context-Aware Simultaneous Translators (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing SiMT systems operate on a sentence level, disregarding the context established by previous sentences or the broader context implied by previous words. |
| Approach: | They show that open-source LLMs perform on par with or better than some state-of-the-art baselines in simultaneous machine translation tasks, zero-shot. |
| Outcome: | The proposed models perform on par with or better than state-of-the-art baselines in simultaneous machine translation tasks, zero-shot. |
Fast, Effective, and Self-Supervised: Transforming Masked Language Models into Universal Lexical and Sentence Encoders (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing studies have shown that pretrained Masked Language Models are not effective as universal lexical and sentence encoders off-the-shelf, i.e., without further task-specific fine-tuning on NLI, sentence similarity, or paraphrasing tasks using annotated task data. |
| Approach: | They propose a contrastive learning technique which turns pretrained MLMs into effective universal lexical and sentence encoders without additional data. |
| Outcome: | The proposed technique can turn MLMs into effective universal lexical and sentence encoders even without additional data. |
MAGNET: Augmenting Generative Decoders with Representation Learning and Infilling Capabilities (2025.acl-long)
Copied to clipboard
| Challenge: | Decoder-only large language models are increasingly being adapted for bidirectional modeling . however, their reliance on causal attention restricts their effectiveness in tasks that require understanding of bidirectional context. |
| Approach: | They propose a method to adapt decoder-only large language models to generate robust representations and infill missing text spans. |
| Outcome: | The proposed method surpasses strong decoders on token-level and sentence-level representation learning tasks and generates contextually appropriate text infills without excessive repetition of words or phrases. |
Causal2Vec: Improving Decoder-only LLMs as Embedding Models through a Contextual Token (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods modify attention mechanism to be bidirectional, undermining LLMs’ ability to extract semantic information acquired during pre-training. |
| Approach: | They propose a general-purpose embedding model that pre-encodes input text into a single Contextual token and then prepends it to the LLM's input sequence. |
| Outcome: | The proposed model improves performance of decoder-only large language models without altering their architectures or introducing significant computational overhead. |