SYNLAT: Syntax-Aligned Text-Latent Compression for Chain-of-Thought Reasoning
On the map: LLM reasoning → Test-time compute and efficient thinking also fits: Prompting and chain-of-thought reasoning
Abstract (arXiv)
Long chain-of-thought (CoT) traces impose substantial output-token costs. Under constrained budgets, compression must preserve answer-critical information, making boundary placement central. Token-level and fixed-length boundaries can fragment coherent spans such as phrases, formulas, and local derivations, whereas step-level boundaries can bind content requiring different compression actions. We introduce SynLat, a text-latent CoT framework that aligns compression boundaries with syntactic structure through non-overlapping Syntax-Aligned Units (SAUs). An answer-conditioned Teacher constructs progressive KEEP/LATENT targets for a single compression-conditioned Student, which generates mixed reasoning from only the question and requested compression level at inference. Across two Qwen3 Student scales, Standard-CoT and Long-CoT groups, and three compression levels, SynLat matches or exceeds the strongest evaluated baseline in all 12 task-group aggregates and strictly leads in 11 under the reported achieved-CR selection protocol. Overall gains reach 3.6/2.6 points at MEDIUM and 7.0/5.5 points at HIGH for Qwen3-8B/14B, with larger advantages under stronger compression, particularly on Long-CoT groups.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 23 of its 24 references that are in the map. First posted on arXiv 2 Oct 2026.
Builds on
23 of 24 references.
Closest ancestors · 9
Same direction, newest first
- 2026A*-Thought-V2: Efficient Latent Reasoning via Geometric Dynamics of LLM7 Sep
- 2026Why Struggle with Continuous Latents? Interpretable Discrete Latent Reasoning via Rendered Compression29 Jun
- 2026Selective Latent Thinking: Adaptive Compression of LLM Reasoning Chains25 May
- 2026SPOT: Span-level Pause-of-Thought for Efficient and Interpretable Latent Reasoning in Large Language Models6 Mar
- 2026Towards Efficient Large Language Reasoning Models via Extreme-Ratio Chain-of-Thought Compression9 Feb
- 2025Making Slow Thinking Faster: Compressing LLM Chain-of-Thought via Step Entropy5 Aug
- 2025R1-Compress: Long Chain-of-Thought Compression via Chunk Compression and Search22 May
- 2025TokenSkip: Controllable Chain-of-Thought Compression in LLMs17 Feb
- 2025Token Assorted: Mixing Latent and Text Tokens for Improved Language Model Reasoning5 Feb
Other references · 14
- 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
- 2025Qwen3 Technical Reportcited by 8,820
- 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,371
- 2017Neural Discrete Representation Learningcited by 4,712
- 2023Let's Verify Step by Stepcited by 3,789
- 2024OlympiadBench: A Challenging Benchmark for Promoting AGI with Olympiad-Level Bilingual Multimodal Scientific Problemscited by 1,136
- 2019SpanBERT: Improving Pre-training by Representing and Predicting Spanscited by 834
- 2024Training Large Language Models to Reason in a Continuous Latent Spacecited by 797
- 2021Did Aristotle Use a Laptop? A Question Answering Benchmark with Implicit Reasoning Strategiescited by 743
- 2020ProofWriter: Generating Implications, Proofs, and Abductive Statements over Natural Languagecited by 324
- 2025CODI: Compressing Chain-of-Thought into Continuous Space via Self-Distillationcited by 225
- 2018Constituency Parsing with a Self-Attentive Encodercited by 199
2 more
- 2025SIM-CoT: Supervised Implicit Chain-of-Thoughtcited by 70
- 2019Neural Extractive Text Summarization with Syntactic Compressioncited by 37
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2026RuPLaR : Efficient Latent Compression of LLM Reasoning Chains with Rule-Based Priors From Multi-Step to One-Step
- 2026Chain Of Thought Compression: A Theoretical Analysis
- 2026SCA: Segment-Wise CoT Compression with Answer Alignment
- 2026Memory Augmentation Unlocks Efficient Chain-of-Thought Reasoning
- 2026CtrlCoT: Dual-Granularity Chain-of-Thought Compression for Controllable Reasoning
- 2025Activation Steering for Chain-of-Thought Compression
- 2026Zipping the Thought: When and How Compressed Reasoning Data Works in LLM Post-Training
- 2026When Compression Helps and When It Hurts: Condition-Aware Analysis of Chain-of-Thought Distillation