What Really Counts? Examining Step and Token Level Attribution in Multilingual CoT Reasoning
On the map: LLM reasoning → Prompting and chain-of-thought reasoning
Abstract (arXiv)
This study investigates the attribution patterns underlying Chain-of-Thought (CoT) reasoning in multilingual LLMs. While prior works demonstrate the role of CoT prompting in improving task performance, there are concerns regarding the faithfulness and interpretability of the generated reasoning chains. To assess these properties across languages, we applied two complementary attribution methods--ContextCite for step-level attribution and Inseq for token-level attribution--to the Qwen2.5 1.5B-Instruct model using the MGSM benchmark. Our experimental results highlight key findings such as: (1) attribution scores excessively emphasize the final reasoning step, particularly in incorrect generations; (2) structured CoT prompting significantly improves accuracy primarily for high-resource Latin-script languages; and (3) controlled perturbations via negation and distractor sentences reduce model accuracy and attribution coherence. These findings highlight the limitations of CoT prompting, particularly in terms of multilingual robustness and interpretive transparency.
Key facts. Cited by 1 paper in the map (data through 14 Oct 2026). Placed in Prompting and chain-of-thought reasoning, one of 17 directions in LLM reasoning. Builds on 19 of its 29 references that are in the map. First posted on arXiv 19 Nov 2025.
Builds on
19 of 29 references.
Closest ancestors · 2
Same direction, newest first
- 2023Analyzing Chain-of-Thought Prompting in Large Language Models via Gradient-based Feature Attributions25 Jul
- 2023Measuring Faithfulness in Chain-of-Thought Reasoning17 Jul
Other references · 17
- 2023LLaMA: Open and Efficient Foundation Language Modelscited by 14,132
- 2021Training Verifiers to Solve Math Word Problemscited by 10,109
- 2025DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learningcited by 6,330
- 2024Qwen2.5 Technical Reportcited by 4,942
- 2014Neural Machine Translation by Jointly Learning to Align and Translatecited by 4,548
- 2017Axiomatic Attribution for Deep Networkscited by 3,384
- 2022Self-Instruct: Aligning Language Models with Self-Generated Instructionscited by 2,724
- 2013Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Mapscited by 2,371
- 2022Language Models are Multilingual Chain-of-Thought Reasonerscited by 530
- 2024LiveBench: A Challenging, Contamination-Limited LLM Benchmarkcited by 262
- 2023Efficient Guided Generation for Large Language Modelscited by 200
- 2023Do All Languages Cost the Same? Tokenization in the Era of Commercial Language Modelscited by 192
5 more
- 2022Attributed Question Answering: Evaluation and Modeling for Attributed Large Language Modelscited by 112
- 2024ContextCite: Attributing Model Generation to Contextcited by 98
- 2023A Survey of Large Language Models Attributioncited by 71
- 2021Specializing Multilingual Language Models: An Empirical Studycited by 21
- 2024Prompting with Phonemes: Enhancing LLMs' Multilinguality for Non-Latin Script Languagescited by 6
Built on by
Citing papers per year: 2026: 1.
1 paper.
Similar, not cited
Close in wording, no citation either way
- 2023Stress Testing Chain-of-Thought Prompting for Large Language Models
- 2025A Comprehensive Evaluation of Multilingual Chain-of-Thought Reasoning: Performance, Consistency, and Faithfulness Across Languages
- 2022Towards Understanding Chain-of-Thought Prompting: An Empirical Study of What Matters
- 2025Is Chain-of-Thought Reasoning of LLMs a Mirage? A Data Distribution Lens
- 2025Long Chain-of-Thought Reasoning Across Languages
- 2023Towards Better Chain-of-Thought Prompting Strategies: A Survey
- 2024Unveiling the Statistical Foundations of Chain-of-Thought Prompting Methods
- 2025Unveiling the Key Factors for Distilling Chain-of-Thought Reasoning