Research map / Architectures and efficiency / RNNs and normalization
Parallel-Probe: Towards Efficient Parallel Thinking via 2D Probing
Tong Zheng, Chengsong Huang, Runpeng Dai, Yun He, Rui Liu, Xin Ni, Huiwen Bao, Kaishen Wang, Hongtu Zhu, Jiaxin Huang, Furong Huang, Heng Huang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add global 2D probing signals (vs local SAC early stopping) (ablation) | Parallel-Probe (Qwen3-0.6B) · AIME24/AIME25 average (ablation) | accuracy: 22.4 -> 25.8 | Improved |
| add global 2D probing signals (vs local SAC early stopping) (ablation) | Parallel-Probe (Qwen3-0.6B) · AIME24/AIME25 average (ablation) | sequential tokens: +33.7% | Improved |
| add deviation-based branch pruning (ablation) | Parallel-Probe (Qwen3-0.6B) · AIME24/AIME25 average (ablation) | accuracy: 23.9 -> 25.8 | No clear effect |
| add deviation-based branch pruning (ablation) | Parallel-Probe (Qwen3-0.6B) · AIME24/AIME25 average (ablation) | total tokens: +14.7% | Improved |
| add consensus-based early stopping (ablation) | Parallel-Probe (Qwen3-0.6B) · AIME24/AIME25 average (ablation) | accuracy: 25.5 -> 25.8 | No clear effect |
| add consensus-based early stopping (ablation) | Parallel-Probe (Qwen3-0.6B) · AIME24/AIME25 average (ablation) | sequential tokens: +13.1% | Improved |
| add warmup stage (ablation) | Parallel-Probe (Qwen3-0.6B) · AIME24/AIME25 average (ablation) | accuracy: 23.5 -> 25.8 | Improved |
| add warmup stage (ablation) | Parallel-Probe (Qwen3-0.6B) · AIME24/AIME25 average (ablation) | total tokens: -19.2% | Hurt |
| replace SC@64 with Parallel-Probe | Qwen3-8B · AIME24/AIME25/HMMT25 average | accuracy: 68.6 -> 68.5 | No clear effect |
| replace SC@64 with Parallel-Probe | Qwen3-0.6B · AIME24/AIME25/HMMT25 average | sequential tokens: -35.8% | Improved |
| replace SC@64 with Parallel-Probe | Qwen3-1.7B · AIME24/AIME25/HMMT25 average | total tokens: -25.8% | Improved |
| apply per-trajectory SAC early stopping to SC@64 | Qwen3-8B · AIME24/AIME25/HMMT25 average | accuracy: 68.6 -> 63.2 | Hurt |
Closest papers by meaning
- ThreadWeaver: Adaptive Threading for Efficient Parallel Reasoning in Language Models (ICML 2026 · similarity 0.89)
- Let Me Think! A Long Chain of Thought Can Be Worth Exponentially Many Short Ones (NeurIPS 2025 · similarity 0.88)
- Parallel-R1: Towards Parallel Thinking via Reinforcement Learning (ICLR 2026 · similarity 0.88)
- Continuous Chain of Thought Enables Parallel Exploration and Reasoning (ICLR 2026 · similarity 0.87)
- SCALE: Selective Resource Allocation for Overcoming Performance Bottlenecks in Mathematical Test-time Scaling (AAAI 2026 · similarity 0.86)
- Visual Para-Thinker: Divide-and-Conquer Reasoning for Visual Comprehension (ICML 2026 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
