mlsanityData to 14 Oct 2026

Enhancing Low-Resource Language Reasoning via High-Resource Language Feature Transfer

arXiv 2608.30462 ↗First posted 31 Aug 2026cs.CL cs.AIcited by 0

On the map: LLM reasoning → Reasoning benchmarks and evaluation

Abstract (arXiv)

Large language models exhibit substantial performance variation across languages, even when solving semantically equivalent tasks. Existing analyses often treat this phenomenon as an observational disparity caused by differences in pretraining data, tokenization, or benchmark coverage. We study a complementary hypothesis: high-resource languages (HRLs) may more reliably elicit latent computations useful for task-specific (i.e. mathematical) reasoning, while lower-resource languages (LRLs) may under-activate those computations despite expressing the same task. To test this hypothesis, we introduce a mechanistic intervention framework for identifying and transferring task-relevant sparse latent features across languages. Using sparse autoencoders over residual-stream activations, we isolate features enriched in successful HRL task-specific reasoning while filtering out source-language and generic-generation features. We then construct steering directions from these features and inject them during LRL inference. The resulting interventions test whether the selected features are functionally involved in the observed reasoning gap: suppressing them should impair source-language reasoning, while activating them should partially recover target-language reasoning beyond random and non-task controls. Our framework reframes some cross-lingual reasoning gaps as failures of mechanism elicitation rather than capability absence, and offers a causally testable route to feature-mediated transfer without translation, fine-tuning, or changing the user-facing language.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning. Builds on 24 of its 51 references that are in the map. First posted on arXiv 31 Aug 2026.

Builds on

24 of 51 references.

Closest ancestors · 3

Same direction, newest first

  1. 2025Why Do Multilingual Reasoning Gaps Emerge in Reasoning Language Models?31 Oct
  2. 2025MMLU-ProX: A Multilingual Benchmark for Advanced Large Language Model Evaluation13 Mar
  3. 2024LinguaLIFT: An Effective Two-stage Instruction Tuning Framework for Low-Resource Language Reasoning17 Dec

Other references · 21

  1. 2024Qwen2.5 Technical Reportcited by 4,942
  2. 2023Let's Verify Step by Stepcited by 3,789
  3. 2019HuggingFace's Transformers: State-of-the-art Natural Language Processingcited by 2,218
  4. 2024Gemma 2: Improving Open Language Models at a Practical Sizecited by 1,998
  5. 2022BLOOM: A 176B-Parameter Open-Access Multilingual Language Modelcited by 1,984
  6. 2023Sparse Autoencoders Find Highly Interpretable Features in Language Modelscited by 1,420
  7. 2023Representation Engineering: A Top-Down Approach to AI Transparencycited by 1,336
  8. 2023Steering Llama 2 via Contrastive Activation Additioncited by 1,085
  9. 2022Toy Models of Superpositioncited by 971
  10. 2023Steering Language Models With Activation Engineeringcited by 922
  11. 2022Language Models are Multilingual Chain-of-Thought Reasonerscited by 530
  12. 2024Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Modelscited by 435
9 more
  1. 2024Gemma Scope: Open Sparse Autoencoders Everywhere All At Once on Gemma 2cited by 394
  2. 2024Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse Autoencoderscited by 300
  3. 2024Programming Refusal with Conditional Activation Steeringcited by 206
  4. 2023CulturaX: A Cleaned, Enormous, and Multilingual Dataset for Large Language Models in 167 Languagescited by 174
  5. 2024Personalized Steering of Large Language Models: Versatile Steering Vectors Through Bi-directional Preference Optimizationcited by 118
  6. 2024Improving Steering Vectors by Targeting Sparse Autoencoder Featurescited by 92
  7. 2024Steering Knowledge Selection Behaviours in LLMs via SAE-Based Representation Engineeringcited by 63
  8. 2023GlotLID: Language Identification for Low-Resource Languagescited by 39
  9. 2025MMATH: A Multilingual Benchmark for Mathematical Reasoningcited by 17

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025I Have Covered All the Bases Here: Interpreting Reasoning Features in Large Language Models via Sparse Autoencoders
  2. 2025Reasoning Transfer for an Extremely Low-Resource and Endangered Language: Bridging Languages Through Sample-Efficient Language Understanding
  3. 2026Do Sparse Autoencoders Identify Reasoning Features in Language Models?
  4. 2025When Less Language is More: Language-Reasoning Disentanglement Makes LLMs Better Multilingual Reasoners
  5. 2025Language Matters: How Do Multilingual Input and Reasoning Paths Affect Large Reasoning Models?
  6. 2026A Systematic Study of Small Language Models on Abstract Reasoning Tasks
  7. 2024Language Models are Hidden Reasoners: Unlocking Latent Reasoning Capabilities via Self-Rewarding
  8. 2026Thinking in a Low-Resource Language: What SFT Builds, What RL Fixes, What Accuracy Cannot See