TUDUM: A Turkish-Thinking Reasoning Pipeline for Qwen3.5-27B
On the map: LLM reasoning → Test-time compute and efficient thinking
Abstract (arXiv)
This paper presents TUDUM (T"urkcce D"ucs"unen "Uretken Model), a project pipeline for adapting a Qwen-family 27B thinking model toward Turkish reasoning. The central problem is not only to answer Turkish prompts in Turkish, but to make the explicit reasoning trace itself Turkish. A thinking model may translate a Turkish prompt into an English-centered internal or visible scratchpad, solve the problem mostly in English, and only localize the final answer. TUDUM instead treats the generated <think>...</think> block as a trainable behavior. The pipeline starts from the project base checkpoint unsloth/Qwen3.5-27B, applies supervised fine-tuning (SFT) on 15,991 Turkish reasoning examples using LoRA adapters, and then applies GRPO-family reinforcement learning on a proxy-filtered Turkish mathematics environment. The results are mixed. SFT made the model shorter and more consistently Turkish in its reasoning behavior, with large reductions in average response length and thinking exhaustion, but reduced benchmark accuracy. RL recovered some mathematical performance, especially AIME24 at the best early checkpoint, yet did not uniformly improve all benchmarks and did not exceed the base model on the reported Macro-6 average. The contribution is therefore best framed as a technically honest Turkish-thinking reasoning pipeline and evaluation, not as a claim of state-of-the-art Turkish reasoning. The released step-50 model is publicly available.
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 16 of its 20 references that are in the map. First posted on arXiv 2 Jul 2026.
Builds on
16 of 20 references.
References · 16
- 2022Training language models to follow instructions with human feedbackcited by 15,712
- 2017Proximal Policy Optimization Algorithmscited by 14,441
- 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,084
- 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,302
- 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
- 2025Qwen3 Technical Reportcited by 8,195
- 2021Evaluating Large Language Models Trained on Codecited by 8,053
- 2020Measuring Massive Multitask Language Understandingcited by 7,768
- 2023GPQA: A Graduate-Level Google-Proof Q&A Benchmarkcited by 2,748
- 2025DAPO: An Open-Source LLM Reinforcement Learning System at Scalecited by 2,678
- 2023Instruction-Following Evaluation for Large Language Modelscited by 1,143
- 2025DeepSeek-V3.2: Pushing the Frontier of Open Large Language Modelscited by 755
4 more
- 2022Language Models are Multilingual Chain-of-Thought Reasonerscited by 518
- 2026Kimi K2.5: Visual Agentic Intelligencecited by 449
- 2026GDPO: Group reward-Decoupled Normalization Policy Optimization for Multi-reward RL Optimizationcited by 175
- 2025Do Multilingual LLMs Think In English?cited by 75
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2025AdapThink: Adaptive Thinking Preferences for Reasoning Language Model
- 2024Bridging the Bosphorus: Advancing Turkish Large Language Models through Strategies for Low-Resource Language Adaptation and Benchmarking
- 2025CORE: Concept-Oriented Reinforcement for Bridging the Definition-Application Gap in Mathematical Reasoning
- 2025Thinking About Thinking: Evaluating Reasoning in Post-Trained Language Models
- 2024TurkishMMLU: Measuring Massive Multitask Language Understanding in Turkish
- 2025Thinker: Learning to Think Fast and Slow
- 2025Towards Large Reasoning Models: A Survey of Reinforced Reasoning with Large Language Models
- 2024Multi-Step Reasoning with Large Language Models, a Survey