mlsanityData to 7 Oct 2026

The Impact of Language Mixing on Bilingual LLM Reasoning

arXiv 2507.15849 ↗First posted 21 Jul 2025cs.CL cs.AI cs.LGcited by 11

On the map: LLM reasoning → Reasoning benchmarks and evaluation

Abstract (arXiv)

Proficient multilingual speakers often intentionally switch languages in the middle of a conversation. Similarly, recent reasoning-focused bilingual large language models (LLMs) with strong capabilities in both languages exhibit language mixing-alternating languages within their chain of thought. Discouraging this behavior in DeepSeek-R1 was found to degrade accuracy, suggesting that language mixing may benefit reasoning. In this work, we study language switching in Chinese-English bilingual reasoning models. We identify reinforcement learning with verifiable rewards (RLVR) as the critical training stage that leads to language mixing. We show that language mixing can enhance reasoning: enforcing monolingual decoding reduces accuracy by 5.6 percentage points on MATH500. Additionally, a lightweight probe can be trained to predict whether a potential language switch would benefit or harm reasoning, and when used to guide decoding, increases accuracy by 2.92 percentage points. Our findings suggest that language mixing is not merely a byproduct of multilingual training, but is a strategic reasoning behavior.

Key facts. Cited by 11 papers in the map (data through 7 Oct 2026). Placed in Reasoning benchmarks and evaluation, one of 17 directions in LLM reasoning. Builds on 18 of its 41 references that are in the map. First posted on arXiv 21 Jul 2025.

Builds on

18 of 41 references.

Closest ancestors · 1

Same direction, newest first

  1. 2024Beyond English-Centric LLMs: What Language Do Multilingual Language Models Think in?20 Aug

Other references · 17

  1. 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,146
  2. 2024Qwen2.5 Technical Reportcited by 4,713
  3. 2023Let's Verify Step by Stepcited by 3,598
  4. 2024Qwen2 Technical Reportcited by 2,076
  5. 2024DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Modelcited by 1,093
  6. 2024Tulu 3: Pushing Frontiers in Open Language Model Post-Trainingcited by 988
  7. 2017Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithmcited by 723
  8. 2023Efficient and Effective Text Encoding for Chinese LLaMA and Alpacacited by 302
  9. 2022Draft, Sketch, and Prove: Guiding Formal Theorem Provers with Informal Proofscited by 278
  10. 2025Logic-RL: Unleashing LLM Reasoning with Rule-Based Reinforcement Learningcited by 260
  11. 2025SFT or RL? An Early Investigation into Training R1-Like Reasoning Large Vision-Language Modelscited by 209
  12. 2024Understanding and Mitigating Language Confusion in LLMscited by 75
5 more
  1. 2025Thinking with Generated Imagescited by 71
  2. 2024The Same But Different: Structural Similarities and Differences in Multilingual Language Modelingcited by 27
  3. 2025Tina: Tiny Reasoning Models via LoRAcited by 23
  4. 2025Investigating and Scaling up Code-Switching for Multilingual Language Model Pre-Trainingcited by 15
  5. 2025Enhancing LLM Language Adaption through Cross-lingual In-Context Pre-trainingcited by 4

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 3; 2026: 8.

11 papers.

  1. 2025Gradual Code-Switching as Inference-Time Cross-Lingual Representational Alignment for LLMscited by 3
  2. 2026Language of Thought Shapes Output Diversity in Large Language Modelscited by 3
  3. 2025Tandem Training for Language Modelscited by 2
  4. 2026OLA: Output Language Alignment in Code-Switched LLM Interactionscited by 2
  5. 2026Think Multilingual, Not Harder: A Data-Efficient Framework for Teaching Reasoning Models to Code-Switchcited by 2
  6. 2026Reading Between the Dots: Decoding Hidden Computation across Filler Tokenscited by 2
  7. 2025Language Confusion Gate: Language-Aware Decoding Through Model Self-Distillationcited by 1
  8. 2026Learning to Route Languages for Multilingual Policy Optimizationcited by 1
  9. 2026HiMed: Incentivizing Hindi Reasoning in Medical LLMscited by 0
  10. 2026Tandem Reinforcement Learning with Verifiable Rewardscited by 0
  11. 2026What Does Post-Training Change in Multilingual Reasoning?cited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026LANG: Reinforcement Learning for Multilingual Reasoning with Language-Adaptive Hint Guidance
  2. 2025Beyond English-Centric Training: How Reinforcement Learning Improves Cross-Lingual Reasoning in LLMs
  3. 2025Cross-lingual Collapse: How Language-Centric Foundation Models Shape Reasoning in Large Language Models
  4. 2025When Less Language is More: Language-Reasoning Disentanglement Makes LLMs Better Multilingual Reasoners
  5. 2026Cross-lingual Self-Consistency for Multilingual Reasoning with Language Models
  6. 2025Language Mixing in Reasoning Language Models: Patterns, Impact, and Internal Causes
  7. 2026Efficient Multilingual Reasoning Transfer via Progressive Code-Switching
  8. 2026Do LLMs Need Inherent Reasoning Before Reinforcement Learning? A Study in Korean Self-Correction