{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,24]],"date-time":"2026-09-24T17:59:21Z","timestamp":1790272761880,"version":"4.1.0"},"reference-count":209,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","issue":"10","license":[{"start":{"date-parts":[[2026,10,1]],"date-time":"2026-10-01T00:00:00Z","timestamp":1790812800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/ieeexplore.ieee.org\/Xplorehelp\/downloads\/license-information\/IEEE.html"},{"start":{"date-parts":[[2026,10,1]],"date-time":"2026-10-01T00:00:00Z","timestamp":1790812800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,10,1]],"date-time":"2026-10-01T00:00:00Z","timestamp":1790812800000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans. Pattern Anal. Mach. Intell."],"published-print":{"date-parts":[[2026,10]]},"DOI":"10.1109\/tpami.2026.3701928","type":"journal-article","created":{"date-parts":[[2026,6,9]],"date-time":"2026-06-09T19:52:16Z","timestamp":1781034736000},"page":"12694-12712","source":"Crossref","is-referenced-by-count":1,"title":["Efficient Inference for Large Reasoning Models: A Survey"],"prefix":"10.1109","volume":"48","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-9894-0062","authenticated-orcid":false,"given":"Yue","family":"Liu","sequence":"first","affiliation":[{"name":"National University of Singapore"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0001-3843-7115","authenticated-orcid":false,"given":"Jiaying","family":"Wu","sequence":"additional","affiliation":[{"name":"National University of Singapore"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-8918-6734","authenticated-orcid":false,"given":"Yufei","family":"He","sequence":"additional","affiliation":[{"name":"National University of Singapore"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ruihan","family":"Gong","sequence":"additional","affiliation":[{"name":"Huazhong University of Science and Technology (HUST)","place":["Hubei, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jun","family":"Xia","sequence":"additional","affiliation":[{"name":"Hong Kong University of Science and Technology","place":["Guangzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-8116-0668","authenticated-orcid":false,"given":"Liang","family":"Li","sequence":"additional","affiliation":[{"name":"National University of Singapore"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hongcheng","family":"Gao","sequence":"additional","affiliation":[{"name":"Universities and Colleges Admissions Service","place":["Beijing, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hongyu","family":"Chen","sequence":"additional","affiliation":[{"name":"Universities and Colleges Admissions Service","place":["Beijing, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0003-4027-1366","authenticated-orcid":false,"given":"Baolong","family":"Bi","sequence":"additional","affiliation":[{"name":"Universities and Colleges Admissions Service","place":["Beijing, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-7950-7845","authenticated-orcid":false,"given":"Jiaheng","family":"Zhang","sequence":"additional","affiliation":[{"name":"National University of Singapore"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zhiqi","family":"Huang","sequence":"additional","affiliation":[{"name":"Peking University","place":["Beijing, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-5645-1754","authenticated-orcid":false,"given":"Bryan","family":"Hooi","sequence":"additional","affiliation":[{"name":"National University of Singapore"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2961-8096","authenticated-orcid":false,"given":"Stan Z.","family":"Li","sequence":"additional","affiliation":[{"name":"Westlake University","place":["Hangzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5224-4048","authenticated-orcid":false,"given":"Keqin","family":"Li","sequence":"additional","affiliation":[{"name":"State University of New York","place":["New York, NY, USA"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","article-title":"Introducing ChatGPT","author":"OpenAI","year":"2022"},{"key":"ref2","article-title":"American invitational mathematics examination","author":"Olympiad","year":"2025"},{"key":"ref3","article-title":"Introducing deep research","author":"OpenAI","year":"2025"},{"key":"ref4","article-title":"OpenAI o1 system card","author":"Jaech","year":"2024"},{"key":"ref5","article-title":"OpenAI o3-mini system card","author":"OpenAI","year":"2025"},{"key":"ref6","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-025-09422-z"},{"key":"ref7","article-title":"Kimi k1. 5: Scaling reinforcement learning with LLMs","author":"Team","year":"2025"},{"key":"ref8","article-title":"Kimi k2: Open agentic intelligence","author":"Team","year":"2025"},{"key":"ref9","article-title":"Rethinking bottlenecks in safety fine-tuning of vision language models","author":"Ding","year":"2025"},{"key":"ref10","article-title":"Model compression and efficient inference for large language models: A survey","author":"Wang","year":"2024"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.52202\/085713-1689"},{"key":"ref12","article-title":"DeepSeek-V3 technical report","author":"Liu","year":"2024"},{"key":"ref13","article-title":"A survey on efficient inference for large language models","author":"Zhou","year":"2024"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.emnlp-main.1236"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1016\/j.ins.2026.123939"},{"key":"ref16","article-title":"Chain of draft: Thinking faster by writing less","author":"Xu","year":"2025"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.findings-acl.1274"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2026.findings-acl.649"},{"key":"ref19","article-title":"Making small language models efficient reasoners: Intervention, supervision, reinforcement","author":"Zhang","year":"2025"},{"key":"ref20","article-title":"Fractured chain-of-thought reasoning","author":"Liao","year":"2025"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.52202\/085713-4668"},{"key":"ref22","article-title":"ThinkLess: A training-free inference-efficient method for reducing reasoning redundancy","author":"Li","year":"2025"},{"key":"ref23","article-title":"Plan and budget: Effective and efficient test-time scaling on large language model reasoning","author":"Lin","year":"2025"},{"key":"ref24","article-title":"TRIMR: Verifier-based training-free thinking compression for efficient test-time scaling","author":"Lin","year":"2025"},{"key":"ref25","article-title":"Solar: Scalable optimization of large-scale architecture for reasoning","author":"Li","year":"2025"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v39i23.34608"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.emnlp-main.165"},{"key":"ref28","article-title":"InftyThink: Breaking the length limits of long-context reasoning in large language models","author":"Yan","year":"2025"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.emnlp-main.673"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.acl-long.300"},{"key":"ref31","article-title":"Distilling system 2 into system 1","author":"Yu","year":"2024"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.findings-acl.1289"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.52202\/079017-1441"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.emnlp-industry.160"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2026.findings-acl.697"},{"key":"ref36","article-title":"Optimizing test-time compute via meta reinforcement fine-tuning","author":"Qu","year":"2025"},{"key":"ref37","doi-asserted-by":"publisher","DOI":"10.52202\/085713-2032"},{"key":"ref38","article-title":"Claude 3.7 Sonnet and Claude code","author":"Anthropic","year":"2025"},{"key":"ref39","article-title":"L1: Controlling how long a reasoning model thinks with reinforcement learning","author":"Aggarwal","year":"2025"},{"key":"ref40","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.findings-acl.956"},{"key":"ref41","article-title":"Think smarter not harder: Adaptive reasoning with inference aware optimization","author":"Yu","year":"2025"},{"key":"ref42","article-title":"Long-short chain-of-thought mixture supervised fine-tuning eliciting efficient reasoning in large language models","author":"Yu","year":"2025"},{"key":"ref43","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.emnlp-main.405"},{"key":"ref44","article-title":"Scalable chain of thoughts via elastic reasoning","author":"Xu","year":"2025"},{"key":"ref45","doi-asserted-by":"publisher","DOI":"10.52202\/085713-1608"},{"key":"ref46","article-title":"Accelerating chain-of-thought reasoning: When goal-gradient importance meets dynamic skipping","author":"Zhuang"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.52202\/085713-5760"},{"key":"ref48","doi-asserted-by":"publisher","DOI":"10.52202\/068431-1800"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.52202\/085713-3775"},{"key":"ref50","article-title":"VAPO: Efficient and reliable reinforcement learning for advanced reasoning tasks","author":"Yue","year":"2025"},{"key":"ref51","article-title":"SimpleRL-Zoo: Investigating and taming zero reinforcement learning for open base models in the wild","author":"Zeng","year":"2025"},{"key":"ref52","article-title":"Efficient reinforcement finetuning via adaptive curriculum learning","author":"Shi","year":"2025"},{"key":"ref53","doi-asserted-by":"publisher","DOI":"10.1007\/s11432-025-4665-8"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.emnlp-main.1025"},{"key":"ref55","doi-asserted-by":"publisher","DOI":"10.52202\/085713-4815"},{"key":"ref56","article-title":"Inference scaling laws: An empirical analysis of compute-optimal inference for LLM problem-solving","volume-title":"Proc. 13th Int. Conf. Learn. Representations","author":"Wu","year":"2024"},{"key":"ref57","article-title":"Scaling LLM test-time compute optimally can be more effective than scaling parameters for reasoning","author":"Snell","year":"2024"},{"key":"ref58","article-title":"Think deep, think fast: Investigating efficiency of verifier-free inference-time-scaling methods","author":"Wang","year":"2025"},{"key":"ref59","article-title":"To CoT or not to CoT? Chain-of-thought helps mainly on math and symbolic reasoning","volume-title":"Proc. ICLR","author":"Sprague","year":"2025"},{"key":"ref60","article-title":"RedStar: Does scaling long-CoT data unlock better slow-reasoning systems","author":"Xu","year":"2025"},{"key":"ref61","article-title":"O1 replication journey\u2013Part 3: Inference-time scaling for medical reasoning","author":"Huang","year":"2025"},{"key":"ref62","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v39i23.34663"},{"key":"ref63","article-title":"Do not think that much for 2 3=? On the overthinking of o1-like LLMs","author":"Chen","year":"2024"},{"key":"ref64","article-title":"When more is less: Understanding chain-of-thought length in LLMs","author":"Wu","year":"2025"},{"key":"ref65","article-title":"The danger of overthinking: Examining the reasoning-action dilemma in agentic tasks","author":"Cuadron","year":"2025"},{"key":"ref66","article-title":"Overthink: Slowdown attacks on reasoning LLMs","author":"Kumar","year":"2025"},{"key":"ref67","doi-asserted-by":"publisher","DOI":"10.52202\/085713-3403"},{"key":"ref68","article-title":"Reasoning models can be effective without thinking","author":"Ma","year":"2025"},{"key":"ref69","article-title":"Cothink: Token-efficient reasoning via instruct models guiding reasoning models","author":"Fan","year":"2025"},{"key":"ref70","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.findings-emnlp.394"},{"key":"ref71","article-title":"Efficient reasoning through suppression of self-affirmation reflections in large reasoning models","author":"Liu","year":"2025"},{"key":"ref72","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.findings-naacl.383"},{"key":"ref73","doi-asserted-by":"publisher","DOI":"10.52202\/079017-1025"},{"key":"ref74","article-title":"Reward-guided speculative decoding for efficient LLM reasoning","author":"Liao","year":"2025"},{"key":"ref75","article-title":"FlashThink: An early exit method for efficient reasoning","author":"Jiang","year":"2025"},{"key":"ref76","article-title":"Large reasoning models know how to think efficiently","author":"Xing","year":"2025"},{"key":"ref77","doi-asserted-by":"publisher","DOI":"10.52202\/085713-2623"},{"key":"ref78","article-title":"Not all tokens are what you need in thinking","author":"Yuan","year":"2025"},{"key":"ref79","doi-asserted-by":"publisher","DOI":"10.52202\/085713-4337"},{"key":"ref80","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.emnlp-main.904"},{"key":"ref81","doi-asserted-by":"publisher","DOI":"10.52202\/085713-5246"},{"key":"ref82","article-title":"Activation steering for chain-of-thought compression","author":"Azizi","year":"2025"},{"key":"ref83","article-title":"Uncovering latent chain of thought vectors in large language models","volume-title":"Proc. ICLR Workshop Neural Netw. Weights New Data Modality","author":"Zhang","year":"2025"},{"key":"ref84","article-title":"SelfBudgeter: Adaptive token allocation for efficient LLM reasoning","author":"Li","year":"2025"},{"key":"ref85","article-title":"Can pruning improve reasoning? revisiting long-CoT compression with capability in mind for better reasoning","author":"Zhao","year":"2025"},{"key":"ref86","doi-asserted-by":"publisher","DOI":"10.52202\/085713-4204"},{"key":"ref87","article-title":"DRP: Distilled reasoning pruning with skill-aware step decomposition for efficient large reasoning models","author":"Jiang","year":"2025"},{"key":"ref88","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.findings-emnlp.770"},{"key":"ref89","article-title":"R1-compress: Long chain-of-thought compression via chunk compression and search","author":"Wang","year":"2025"},{"key":"ref90","doi-asserted-by":"publisher","DOI":"10.52202\/085713-3594"},{"key":"ref91","article-title":"Non-myopic generation of language models for reasoning and planning","author":"Ma","year":"2024"},{"key":"ref92","doi-asserted-by":"publisher","DOI":"10.52202\/079017-1740"},{"key":"ref93","article-title":"LIMO: Less is more for reasoning","author":"Ye","year":"2025"},{"key":"ref94","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.emnlp-main.333"},{"key":"ref95","article-title":"Improving mathematical reasoning capabilities of small language models via feedback-driven distillation","author":"Zhu","year":"2024"},{"key":"ref96","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.findings-acl.1068"},{"key":"ref97","article-title":"SmartThinker: Learning to compress and preserve reasoning by step-level length control","author":"He","year":"2025"},{"key":"ref98","article-title":"PREMISE: Scalable and strategic prompt optimization for efficient mathematical reasoning in large models","author":"Yu","year":"2025"},{"key":"ref99","article-title":"Less data less tokens: Multilingual unification learning for efficient test-time reasoning in LLMs","author":"Chen","year":"2025"},{"key":"ref100","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.findings-emnlp.845"},{"key":"ref101","article-title":"ConciseHint: Boosting efficient reasoning via continuous concise hints during generation","author":"Tang"},{"key":"ref102","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2026.findings-acl.1866"},{"key":"ref103","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2026.acl-long.1856"},{"key":"ref104","doi-asserted-by":"publisher","DOI":"10.52202\/085713-5629"},{"key":"ref105","article-title":"Implicit chain of thought reasoning via knowledge distillation","author":"Deng","year":"2023"},{"key":"ref106","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.emnlp-main.36"},{"key":"ref107","article-title":"From explicit cot to implicit CoT: Learning to internalize CoT step by step","author":"Deng","year":"2024"},{"key":"ref108","article-title":"Training large language models to reason in a continuous latent space","author":"Hao","year":"2024"},{"key":"ref109","article-title":"Compressed chain of thought: Efficient reasoning through dense representations","author":"Cheng","year":"2024"},{"key":"ref110","article-title":"Efficient reasoning with hidden thinking","author":"Shen","year":"2025"},{"key":"ref111","article-title":"Token assorted: Mixing latent and text tokens for improved language model reasoning","author":"Su","year":"2025"},{"key":"ref112","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.acl-long.1137"},{"key":"ref113","doi-asserted-by":"publisher","DOI":"10.52202\/085713-0164"},{"key":"ref114","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v40i40.40659"},{"key":"ref115","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.emnlp-main.256"},{"key":"ref116","article-title":"Rethinking LLM-based preference evaluation","author":"Hu","year":"2024"},{"key":"ref117","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.findings-emnlp.358"},{"key":"ref118","doi-asserted-by":"publisher","DOI":"10.52202\/085713-1302"},{"key":"ref119","article-title":"Fast on the easy, deep on the hard: Efficient reasoning via powered length penalty","author":"Ling","year":"2025"},{"key":"ref120","article-title":"ThinkPrune: Pruning long chain-of-thought of LLMs via reinforcement learning","author":"Hou","year":"2025"},{"key":"ref121","article-title":"Do thinking tokens help or trap? Towards more efficient large reasoning model","author":"Ding","year":"2025"},{"key":"ref122","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2026.acl-long.726"},{"key":"ref123","article-title":"AdapThink: Adaptive thinking preferences for reasoning language model","author":"Wan","year":"2025"},{"key":"ref124","article-title":"Think when you need: Self-adaptive chain-of-thought learning","author":"Yang","year":"2025"},{"key":"ref125","article-title":"TINA: Tiny reasoning models via LoRA","author":"Wang","year":"2025"},{"key":"ref126","doi-asserted-by":"publisher","DOI":"10.52202\/085713-0365"},{"key":"ref127","article-title":"Exploring and exploiting the inherent efficiency within large reasoning models for self-guided efficiency enhancement","author":"Zhao","year":"2025"},{"key":"ref128","article-title":"Efficient RL training for reasoning models via length-aware optimization","author":"Yuan","year":"2025"},{"key":"ref129","article-title":"Learn to reason efficiently with adaptive length-based reward shaping","author":"Liu","year":"2025"},{"key":"ref130","doi-asserted-by":"publisher","DOI":"10.52202\/085713-5108"},{"key":"ref131","doi-asserted-by":"publisher","DOI":"10.52202\/085713-0793"},{"key":"ref132","article-title":"Overclocking LLM reasoning: Monitoring and controlling thinking path lengths in LLMs","author":"Eisenstadt","year":"2025"},{"key":"ref133","article-title":"Not all thoughts are generated equal: Efficient LLM reasoning via multi-turn reinforcement learning","author":"Ning","year":"2025"},{"key":"ref134","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.findings-emnlp.927"},{"key":"ref135","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.findings-acl.1125"},{"key":"ref136","article-title":"Walk before you run! concise LLM reasoning via reinforcement learning","author":"Song","year":"2025"},{"key":"ref137","article-title":"Optimizing length compression in large reasoning models","author":"Cheng","year":"2025"},{"key":"ref138","article-title":"AALC: Large language model efficient reasoning via adaptive accuracy-length control","author":"Li","year":"2025"},{"key":"ref139","doi-asserted-by":"publisher","DOI":"10.52202\/085713-3707"},{"key":"ref140","article-title":"Expediting and elevating large language model reasoning via hidden chain-of-thought decoding","author":"Liu"},{"key":"ref141","article-title":"Do LLMs really think step-by-step in implicit reasoning","author":"Yu","year":"2025"},{"key":"ref142","article-title":"Training verifiers to solve math word problems","author":"Cobbe"},{"key":"ref143","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.eacl-short.15"},{"key":"ref144","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2021.naacl-main.168"},{"key":"ref145","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/P17-1015"},{"key":"ref146","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2020.acl-main.92"},{"key":"ref147","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.findings-acl.411"},{"key":"ref148","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.emnlp-main.489"},{"key":"ref149","article-title":"Measuring mathematical problem solving with the math dataset","volume-title":"Proc. NeurIPS","author":"Hendrycks","year":"2021"},{"key":"ref150","article-title":"MetaMath: Bootstrap your own mathematical questions for large language models","author":"Yu"},{"key":"ref151","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.211"},{"key":"ref152","article-title":"GPQA: A graduate-level google-proof Q&A benchmark","volume-title":"Proc. First Conf. Lang. Model.","author":"Rein","year":"2024"},{"key":"ref153","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v34i05.6319"},{"key":"ref154","doi-asserted-by":"publisher","DOI":"10.63317\/28283yn8kzpn"},{"key":"ref155","article-title":"LiveCodeBench: Holistic and contamination free evaluation of large language models for code","author":"Jain"},{"key":"ref156","article-title":"SWE-bench: Can language models resolve real-world github issues?","author":"Jimenez"},{"key":"ref157","article-title":"Language models are greedy reasoners: A systematic formal analysis of chain-of-thought","volume-title":"Proc. ICLR","author":"Saparov","year":"2023"},{"key":"ref158","doi-asserted-by":"publisher","DOI":"10.24963\/ijcai.2020\/501"},{"key":"ref159","article-title":"ReClor: A reading comprehension dataset requiring logical reasoning","volume-title":"Proc. ICLR","author":"Yu","year":"2020"},{"key":"ref160","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/N19-1421"},{"key":"ref161","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/D18-1260"},{"key":"ref162","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2021.acl-long.238"},{"key":"ref163","doi-asserted-by":"publisher","DOI":"10.1162\/tacl_a_00370"},{"key":"ref164","article-title":"Beyond the imitation game: Quantifying and extrapolating the capabilities of language models","author":"Srivastava","year":"2022"},{"key":"ref165","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.findings-acl.824"},{"key":"ref166","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/d18-1259"},{"key":"ref167","doi-asserted-by":"publisher","DOI":"10.1162\/tacl_a_00475"},{"key":"ref168","article-title":"Measuring massive multitask language understanding","author":"Hendrycks","year":"2020"},{"key":"ref169","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.00913"},{"key":"ref170","doi-asserted-by":"publisher","DOI":"10.52202\/068431-0182"},{"key":"ref171","first-page":"50622","article-title":"SciBench: Evaluating college-level scientific problem-solving abilities of large language models","volume-title":"Proc. ICML","author":"Wang","year":"2024"},{"key":"ref172","doi-asserted-by":"publisher","DOI":"10.52202\/079017-3014"},{"key":"ref173","article-title":"Evaluating mathematical reasoning of foundation models in visual contexts","volume-title":"Proc. ICLR","author":"Lu","year":"2024"},{"key":"ref174","article-title":"\u03c4-bench: A benchmark for tool-agent-user interaction in real-world domains","author":"Yao"},{"key":"ref175","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/D19-1259"},{"key":"ref176","doi-asserted-by":"publisher","DOI":"10.56726\/irjmets45213"},{"key":"ref177","doi-asserted-by":"publisher","DOI":"10.1186\/s13000-024-01464-7"},{"key":"ref178","doi-asserted-by":"publisher","DOI":"10.1145\/3630106.3659048"},{"key":"ref179","article-title":"Monitoring monitorability","author":"Guan","year":"2025"},{"key":"ref180","article-title":"Chain-of-thought reasoning in the wild is not always faithful","author":"Arcuschin","year":"2025"},{"key":"ref181","doi-asserted-by":"publisher","DOI":"10.52202\/085713-1654"},{"key":"ref182","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v40i38.40486"},{"key":"ref183","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.254"},{"key":"ref184","article-title":"General purpose verification for chain of thought prompting","author":"Vacareanu","year":"2024"},{"key":"ref185","article-title":"A pragmatic way to measure chain-of-thought monitorability","author":"Emmons","year":"2025"},{"key":"ref186","article-title":"FlipAttack: Jailbreak LLMs via flipping","author":"Liu","year":"2024"},{"key":"ref187","article-title":"Evaluating the paperclip maximizer: Are RL-based language models more likely to pursue instrumental goals","author":"He","year":"2025"},{"key":"ref188","article-title":"Privacy in large language models: Attacks, defenses and future directions","author":"Li"},{"key":"ref189","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.findings-emnlp.185"},{"key":"ref190","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v34i05.6319"},{"key":"ref191","doi-asserted-by":"publisher","DOI":"10.52202\/085713-0977"},{"key":"ref192","doi-asserted-by":"publisher","DOI":"10.52202\/085713-0977"},{"key":"ref193","article-title":"Efficient reasoning via chain of unconscious thought","author":"Gong","year":"2025"},{"key":"ref194","doi-asserted-by":"publisher","DOI":"10.1007\/s13278-025-01428-9"},{"key":"ref195","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2025.emnlp-main.277"},{"key":"ref196","article-title":"Introducing GPT-4.5","author":"OpenAI","year":"2025"},{"key":"ref197","doi-asserted-by":"publisher","DOI":"10.1109\/cvpr52734.2025.00168"},{"key":"ref198","article-title":"Gemini robotics brings AI into the physical world","author":"Google","year":"2025"},{"key":"ref199","article-title":"Nvidia isaac GR00T N1: An open foundation model for humanoid","author":"Nvidia","year":"2025"},{"key":"ref200","doi-asserted-by":"publisher","DOI":"10.1145\/3802463.3802475"},{"key":"ref201","article-title":"LLM4Drive: A survey of large language models for autonomous driving","author":"Yang"},{"key":"ref202","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.findings-emnlp.936"},{"key":"ref203","article-title":"Linear-time sequence modeling with selective state spaces","author":"Gu"},{"key":"ref204","doi-asserted-by":"publisher","DOI":"10.1145\/3787849"},{"key":"ref205","article-title":"Unlocking efficient long-to-short LLM reasoning with model merging","author":"Wu","year":"2025"},{"key":"ref206","article-title":"The Llama 3 herd of models","author":"Grattafiori","year":"2024"},{"key":"ref207","article-title":"RouteLLM: Learning to route LLMs from preference data","volume-title":"Proc. ICLR","author":"Ong","year":"2025"},{"key":"ref208","article-title":"Learning to route LLMs with confidence tokens","author":"Chuang","year":"2025"},{"key":"ref209","article-title":"Confident or seek stronger: Exploring uncertainty-based on-device LLM routing from benchmarking to generalization","author":"Chuang","year":"2025"}],"container-title":["IEEE Transactions on Pattern Analysis and Machine Intelligence"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/34\/11674301\/11555030.pdf?arnumber=11555030","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,24]],"date-time":"2026-09-24T17:01:15Z","timestamp":1790269275000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11555030\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,10]]},"references-count":209,"journal-issue":{"issue":"10"},"URL":"https:\/\/doi.org\/10.1109\/tpami.2026.3701928","relation":{},"ISSN":["0162-8828","2160-9292","1939-3539"],"issn-type":[{"value":"0162-8828","type":"print"},{"value":"2160-9292","type":"electronic"},{"value":"1939-3539","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,10]]}}}