How a Bilingual LM Becomes Bilingual: Tracing Internal Representations with Sparse Autoencoders (2025.findings-emnlp)
Copied to clipboard
Tatsuro Inaba, Go Kamoda, Kentaro Inui, Masaru Isonuma, Yusuke Miyao, Yohei Oseki, Yu Takagi, Benjamin Heinzerling
| Challenge: | Using sparse autoencoders, we explore how bilingual language models develop complex internal representations. |
| Approach: | They employ sparse autoencoders to analyze bilingual language models' internal representations. |
| Outcome: | The proposed method integrates decomposed representations from a fully trained model into a mid-training model. |
Similar Papers
Large Language Models Share Representations of Latent Grammatical Concepts Across Typologically Diverse Languages (2025.naacl-long)
Copied to clipboard
| Challenge: | In the brains of human bilinguals, syntax processing may occur in similar regions for their first and second language, depending on factors like when the second language was learned and language proficiency. |
| Approach: | They propose to use sparse autoencoders to train Llama-3-8B and Aya-23-8B models to train multilingual models that share morphsyntactic representations of grammatical concepts. |
| Outcome: | The proposed model can predict plural verbs in different languages by activating the same plural feature. |
Deciphering Cultural Representations in Large Language Models via Sparse Autoencoders (2026.findings-acl)
Copied to clipboard
| Challenge: | Prior work has identified so-called cultural neurons, but individual neurons are often polysemous, conflating abstract cultural knowledge with surface-level lexical cues due to superposition. |
| Approach: | They apply Sparse Autoencoders to decompose LLM activations into sparse, interpretable feature representations that disentangle culturally selective features. |
| Outcome: | The proposed model disentangles culturally selective features from paraphrasing and task formats, indicating abstraction beyond lexical correlations. |
Unveiling Language-Specific Features in Large Language Models via Sparse Autoencoders (2025.acl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) exhibit impressive abilities in various domains such as text generation, instruction following, and reasoning. |
| Approach: | They propose a method to decompose the activations of Large Language Models into a sparse linear combination of SAE features. |
| Outcome: | The proposed method shows that some features are strongly related to specific languages, while others are unaffected by ablating them. |
A Survey on Sparse Autoencoders: Interpreting the Internal Mechanisms of Large Language Models (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Sparse Autoencoders (SAEs) can disentangle complex features into more interpretable components. |
| Approach: | They propose to use Sparse Autoencoders to disentangle LLM features into more interpretable components. |
| Outcome: | The proposed method disentangles complex features into more interpretable components. |
Second Language Acquisition of Neural Language Models (2023.findings-acl)
Copied to clipboard
| Challenge: | a recent study examined the cross-lingual transferability of neural language models . previous studies focused on their first language acquisition . |
| Approach: | They propose to pretrain bilingual LMs with a scenario similar to human L2 acquisition . they find that pretraining accelerated their linguistic generalization in L2 . |
| Outcome: | The results show that pretraining bilingual LMs accelerates their linguistic generalizations . the results clarify their (non-)human-like L2 acquisition in particular aspects . |
Middle-Layer Representation Alignment for Cross-Lingual Transfer in Fine-Tuned LLMs (2025.acl-long)
Copied to clipboard
| Challenge: | Effective cross-lingual transfer is hindered by performance gaps and the scarcity of fine-tuning data in many languages. |
| Approach: | They propose a middle-layer alignment objective integrated into task-specific training to improve cross-lingual transfer across languages. |
| Outcome: | The proposed method improves cross-lingual transfer to lower-resource languages and can be merged with existing modules without full re-training. |
On the Acquisition of Shared Grammatical Representations in Bilingual Language Models (2025.acl-long)
Copied to clipboard
| Challenge: | Crosslingual transfer is crucial to contemporary language models’ multilingual capabilities, but how it occurs is not well understood. |
| Approach: | They use structural priming to study grammatical representations in humans by controlling for training data quantity and language exposure. |
| Outcome: | The proposed model is able to learn a language in two languages and has a higher likelihood of learning a prepositional object (PO) dative sentence than a double object (DO) . |
LinguaLens: Towards Interpreting Linguistic Mechanisms of Large Language Models via Sparse Auto-Encoder (2025.emnlp-main)
Copied to clipboard
| Challenge: | Prior research on linguistic mechanisms of large language models is limited by coarse granularity, limited analysis scale, and narrow focus. |
| Approach: | They propose a framework for analyzing the linguistic mechanisms of large language models based on Sparse Auto-Encoders. |
| Outcome: | The proposed framework extracts Chinese and English linguistic features across four dimensions . it uncovers intrinsic representations of linguistic knowledge in LLMs and can control outputs . |
Language Lives in Sparse Dimensions: Toward Interpretable and Efficient Multilingual Control for Large Language Models (2026.eacl-long)
Copied to clipboard
| Challenge: | Prior studies show that large language models map multilingual content into English-aligned representations at intermediate layers before projecting them back into target-language token spaces in the later layers. |
| Approach: | They propose a method to identify and manipulate dimensions that are sparse and sparsity-based . they propose to use as few as 50 sentences of either parallel or monolingual data to manipulate these dimensions . |
| Outcome: | Experiments on a multilingual generation control task show the interpretability of these dimensions. |
Understanding Cross-Lingual Alignment—A Survey (2024.findings-acl)
Copied to clipboard
| Challenge: | Cross-lingual alignment is the meaningful similarity of representations across languages in multilingual language models. |
| Approach: | They propose a taxonomy of methods to improve cross-lingual alignment . they argue that an effective trade-off between language-neutral and language-specific information is key . |
| Outcome: | The proposed methods can be applied to encoder models and encoder-decoder-only models . they show that language-neutral and language-specific information is key . |