arXiv cs.CL· Juseon Do, Sangmin Bae, Jeesu Jung, Hwanjun Song·· 11 小时前AI 评分34
TAC-CoT：通过交错显式 Token 实现工具赋能的隐式思维链
TAC-CoT: Tool-Empowered Implicit Chain-of-Thought via Interleaving Explicit Tokens
AI 导读
TAC-CoT 是一个工具赋能的隐式 CoT 框架，用紧凑的潜在推理片段替代显式推理文本，仅在工具调用、观察结果和最终答案处保留文本 token。在智能体数学与问答基准上，其准确率最高提升 8.8 分，模型前向传播次数最多减少 6.8 倍。分析显示该框架在单工具和多工具场景下均提升了工具使用质量。
正文 · 原文
Abstract:LLM-based agentic systems are increasingly combined with CoT reasoning to solve tasks that require accurate reasoning, external knowledge, and multi-step deliberation. Yet, explicit CoT incurs substantial generation overhead, while implicit CoT, which replaces textual rationales with compact continuous tokens, does not naturally support the explicit communication interface required for agentic tool-use. We propose TAC-CoT, a tool-empowered adaptive implicit CoT framework that replaces explicit reasoning traces with compact latent reasoning segments while retaining textual tokens only for tool calls, observations, and final answers. This interleaved design enables agentic tool-use without verbalizing every intermediate reasoning step. Experiments on agentic math and QA benchmarks show that TAC-CoT improves accuracy by up to 8.8 points while reducing model forward passes by up to 6.8x. Furthermore, analyses show that TAC-CoT improves tool-use quality in both single- and multi-tool settings.
| Comments: | EMNLP 2026 Main Long |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2611.01926 [cs.CL] |
| (or arXiv:2611.01926v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2611.01926 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Juseon Do [view email]
[v1]
Sun, 1 Nov 2026 14:05:15 UTC (636 KB)
来源：arXiv cs.CL · arxiv.org