![[2602.15038] Indic-TunedLens: Interpreting Multilingual Models in Indian Languages](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.15038] Indic-TunedLens: Interpreting Multilingual Models in Indian Languages
Summary
The paper introduces Indic-TunedLens, an interpretability framework designed for multilingual models in Indian languages, enhancing cross-lingual interpretability and performance on low-resource languages.
Why It Matters
As multilingual large language models are increasingly used in India, understanding their interpretability in diverse linguistic contexts is crucial. Indic-TunedLens addresses the limitations of existing tools that are primarily English-centric, providing a necessary framework for better model understanding and performance in Indian languages, which are often morphologically rich and underrepresented in AI research.
Key Takeaways
- Indic-TunedLens improves interpretability for multilingual models in Indian languages.
- The framework adjusts hidden states to align with target language outputs.
- Significant performance improvements observed on the MMLU benchmark for low-resource languages.
- Addresses the need for tools tailored to linguistically diverse regions.
- Enhances understanding of layer-wise semantic encoding in multilingual transformers.
Computer Science > Computation and Language arXiv:2602.15038 (cs) [Submitted on 29 Jan 2026] Title:Indic-TunedLens: Interpreting Multilingual Models in Indian Languages Authors:Mihir Panchal, Deeksha Varshney, Mamta, Asif Ekbal View a PDF of the paper titled Indic-TunedLens: Interpreting Multilingual Models in Indian Languages, by Mihir Panchal and 3 other authors View PDF HTML (experimental) Abstract:Multilingual large language models (LLMs) are increasingly deployed in linguistically diverse regions like India, yet most interpretability tools remain tailored to English. Prior work reveals that LLMs often operate in English centric representation spaces, making cross lingual interpretability a pressing concern. We introduce Indic-TunedLens, a novel interpretability framework specifically for Indian languages that learns shared affine transformations. Unlike the standard Logit Lens, which directly decodes intermediate activations, Indic-TunedLens adjusts hidden states for each target language, aligning them with the target output distributions to enable more faithful decoding of model representations. We evaluate our framework on 10 Indian languages using the MMLU benchmark and find that it significantly improves over SOTA interpretability methods, especially for morphologically rich, low resource languages. Our results provide crucial insights into the layer-wise semantic encoding of multilingual transformers. Our model is available at this https URL. Our code is availabl...
