Sliding-window transformers without positional encodings are Turing complete because the sliding window breaks permutation symmetry and suffices to simulate Post machines via a constant-size histogram state.
hub Canonical reference
Towards Large Reasoning Models: A Survey of Reinforced Reasoning with Large Language Models
Canonical reference. 100% of citing Pith papers cite this work as background.
51
Pith papers citing it
17
external citations · Pith
Background
100% of classified citations
abstract
Language has long been conceived as an essential tool for human reasoning. The breakthrough of Large Language Models (LLMs) has sparked significant research interest in leveraging these models to tackle complex reasoning tasks. Researchers have moved beyond simple autoregressive token generation by introducing the concept of "thought" -- a sequence of tokens representing intermediate steps in the reasoning process. This innovative paradigm enables LLMs' to mimic complex human reasoning processes, such as tree search and reflective thinking. Recently, an emerging trend of learning to reason has applied reinforcement learning (RL) to train LLMs to master reasoning processes. This approach enables the automatic generation of high-quality reasoning trajectories through trial-and-error search algorithms, significantly expanding LLMs' reasoning capacity by providing substantially more training data. Furthermore, recent studies demonstrate that encouraging LLMs to "think" with more tokens during test-time inference can further significantly boost reasoning accuracy. Therefore, the train-time and test-time scaling combined to show a new research frontier -- a path toward Large Reasoning Model. The introduction of OpenAI's o1 series marks a significant milestone in this research direction. In this survey, we present a comprehensive review of recent progress in LLM reasoning. We begin by introducing the foundational background of LLMs and then explore the key technical components driving the development of large reasoning models, with a focus on automated data construction, learning-to-reason techniques, and test-time scaling. We also analyze popular open-source projects at building large reasoning models, and conclude with open challenges and future research directions.
hub tools
citation-role summary
background 10
citation-polarity summary
roles
background 10polarities
background 10representative citing papers
MirageBackdoor is the first backdoor attack that preserves clean chain-of-thought reasoning in LLMs while steering the final answer to a specific incorrect target under a trigger.
Post-Reasoning boosts LLM accuracy by reversing the usual answer-after-reasoning order, delivering mean relative gains of 17.37% across 117 model-benchmark pairs with zero extra cost.
IE-as-Cache framework repurposes information extraction as a dynamic cognitive cache to improve agentic reasoning accuracy in LLMs on challenging benchmarks.
Metacognitive Behavioral Tuning injects a five-phase structure into LLM reasoning traces to improve accuracy-efficiency on multi-hop QA while reducing trace length and degeneration.
Users treat human delegation for long tasks as a flexible compass but AI delegation as rigid railway tracks due to perceived AI limitations in inference and judgment.
AgentMark watermarks agent planning behaviors with multi-bit identifiers via conditional sampling that preserves utility and works on black-box systems.
Hybrid Bayesian-graph LLM agent reaches competitive performance against large models and achieves 67% win rate against humans in controlled Avalon play, outperforming baselines and human teammates.
ToxiREX is a new dataset of 128k Reddit comments in six languages with hierarchical annotations for implicit toxicity in conversational context based on an existing reasoning schema.
CombEval generates solver-verified combinatorial counting problems to test LLMs on ordered objects, indistinguishable elements, positional constraints, and nested dependencies.
Groups that deliberate on chess puzzles improve their move quality on average, but more probing questions only increase outcome variance unless they focus on concrete answer options.
ThoughtFold applies introspective redundancy detection within correct CoT trajectories to create sub-trajectory spectra, then uses masked preference optimization to penalize redundant explorations, yielding 56% token reduction on DeepSeek-R1-Distill-Qwen-7B while preserving accuracy.
CA-BED uses Bayesian experimental design and simulated conversation trees with LLM likelihoods to optimize multi-turn question selection, reporting 21.8% higher success rates than direct prompting on entity-deduction benchmarks.
ADR generates novel verifiable code tasks via atomic decomposition and recombination, outperforming heuristic baselines in originality, difficulty, and downstream RLVR gains across coding domains.
REALISTA generates semantically coherent adversarial prompts via latent-space optimization over input-dependent editing directions, achieving stronger hallucination elicitation than prior realistic attacks on open-source and reasoning LLMs.
ADMM-Q is a new post-training quantization method using ADMM operator splitting that reduces WikiText-2 perplexity compared to GPTQ on Qwen3-8B across W3A16, W4A8, and W2A4KV4 settings.
Meta-Aligner introduces a meta-learner network that produces dynamic preference weights to enable bidirectional optimization between preferences and LLM policy responses for multi-objective alignment.
LLMs and VLMs exhibit grounding-dependent causal transfer in the OpenLock paradigm, requiring initial context-specific learning unlike humans who transfer structures from the first attempt.
Step-Saliency identifies shallow lock-in and deep decay in reasoning model attention flows, and StepFlow intervention repairs them to improve accuracy across LRMs.
Gome reaches 35.1% any-medal rate on MLE-Bench by mapping reasoning to gradient-based updates, outperforming tree search once models are sufficiently capable.
A rule-aware modular prompt framework enables LLMs to perform structured numeric reasoning on power grid data by separating rules from normalized deviations, improving anomaly detection consistency and reducing token use in IEEE 118-bus tests.
LLMs generate valid solutions to over 70% of AI research problems from parametric memory alone but rediscover the exact published approach less than 19% of the time, with performance limited by cross-domain analogical transfer.
Retrieval-of-Thought organizes prior reasoning into a thought graph for retrieval and reward-guided recombination, reducing output tokens by up to 40% and latency by 82% while preserving accuracy on reasoning benchmarks.
PEER applies GRPO reinforcement learning with a unified process-outcome reward model to structured empathetic reasoning steps on the SER dataset, yielding gains in empathy, strategy alignment, and human-likeness.
citing papers explorer
-
Rethinking the Role of Positional Encoding: Sliding-Window Transformers without PE Remain Turing Complete
Sliding-window transformers without positional encodings are Turing complete because the sliding window breaks permutation symmetry and suffices to simulate Post machines via a constant-size histogram state.
-
MirageBackdoor: A Stealthy Attack that Induces Think-Well-Answer-Wrong Reasoning
MirageBackdoor is the first backdoor attack that preserves clean chain-of-thought reasoning in LLMs while steering the final answer to a specific incorrect target under a trigger.
-
Post Reasoning: Improving the Performance of Non-Thinking Models at No Cost
Post-Reasoning boosts LLM accuracy by reversing the usual answer-after-reasoning order, delivering mean relative gains of 17.37% across 117 model-benchmark pairs with zero extra cost.
-
IE as Cache: Information Extraction Enhanced Agentic Reasoning
IE-as-Cache framework repurposes information extraction as a dynamic cognitive cache to improve agentic reasoning accuracy in LLMs on challenging benchmarks.
-
Metacognitive Behavioral Tuning of Large Language Models for Multi-Hop Question Answering
Metacognitive Behavioral Tuning injects a five-phase structure into LLM reasoning traces to improve accuracy-efficiency on multi-hop QA while reducing trace length and degeneration.
-
Compass vs Railway Tracks: Unpacking User Mental Models for Communicating Long-Horizon Work to Humans vs. AI
Users treat human delegation for long tasks as a flexible compass but AI delegation as rigid railway tracks due to perceived AI limitations in inference and judgment.
-
AgentMark: Utility-Preserving Behavioral Watermarking for Agents
AgentMark watermarks agent planning behaviors with multi-bit identifiers via conditional sampling that preserves utility and works on black-box systems.
-
Bayesian Social Deduction with Graph-Informed Language Models
Hybrid Bayesian-graph LLM agent reaches competitive performance against large models and achieves 67% win rate against humans in controlled Avalon play, outperforming baselines and human teammates.
-
ToxiREX: A Dataset on Toxic REasoning in ConteXt
ToxiREX is a new dataset of 128k Reddit comments in six languages with hierarchical annotations for implicit toxicity in conversational context based on an existing reasoning schema.
-
CombEval: A Framework for Evaluating Combinatorial Counting in Large Language Models
CombEval generates solver-verified combinatorial counting problems to test LLMs on ordered objects, indistinguishable elements, positional constraints, and nested dependencies.
-
DeliChess: A Multi-party Dialogue Dataset for Deliberation in Chess Puzzle Solving
Groups that deliberate on chess puzzles improve their move quality on average, but more probing questions only increase outcome variance unless they focus on concrete answer options.
-
ThoughtFold: Folding Reasoning Chains via Introspective Preference Learning
ThoughtFold applies introspective redundancy detection within correct CoT trajectories to create sub-trajectory spectra, then uses masked preference optimization to penalize redundant explorations, yielding 56% token reduction on DeepSeek-R1-Distill-Qwen-7B while preserving accuracy.
-
CA-BED: Conversation-Aware Bayesian Experimental Design
CA-BED uses Bayesian experimental design and simulated conversation trees with LLM likelihoods to optimize multi-turn question selection, reporting 21.8% higher success rates than direct prompting on entity-deduction benchmarks.
-
Combinatorial Synthesis: Scaling Code RLVR via Atomic Decomposition and Recombination
ADR generates novel verifiable code tasks via atomic decomposition and recombination, outperforming heuristic baselines in originality, difficulty, and downstream RLVR gains across coding domains.
-
REALISTA: Realistic Latent Adversarial Attacks that Elicit LLM Hallucinations
REALISTA generates semantically coherent adversarial prompts via latent-space optimization over input-dependent editing directions, achieving stronger hallucination elicitation than prior realistic attacks on open-source and reasoning LLMs.
-
ADMM-Q: An Improved Hessian-based Weight Quantizer for Post-Training Quantization of Large Language Models
ADMM-Q is a new post-training quantization method using ADMM operator splitting that reduces WikiText-2 perplexity compared to GPTQ on Qwen3-8B across W3A16, W4A8, and W2A4KV4 settings.
-
Meta-Aligner: Bidirectional Preference-Policy Optimization for Multi-Objective LLMs Alignment
Meta-Aligner introduces a meta-learner network that produces dynamic preference weights to enable bidirectional optimization between preferences and LLM policy responses for multi-objective alignment.
-
Grounding Before Generalizing: How AI Differs from Humans in Causal Transfer
LLMs and VLMs exhibit grounding-dependent causal transfer in the OpenLock paradigm, requiring initial context-specific learning unlike humans who transfer structures from the first attempt.
-
Reasoning Fails Where Step Flow Breaks
Step-Saliency identifies shallow lock-in and deep decay in reasoning model attention flows, and StepFlow intervention repairs them to improve accuracy across LRMs.
-
Reasoning as Gradient: Scaling MLE Agents Beyond Tree Search
Gome reaches 35.1% any-medal rate on MLE-Bench by mapping reasoning to gradient-based updates, outperforming tree search once models are sufficiently capable.
-
A Rule-Aware Prompt Framework for Structured Numeric Reasoning in Cyber-Physical Systems
A rule-aware modular prompt framework enables LLMs to perform structured numeric reasoning on power grid data by separating rules from normalized deviations, improving anomaly detection consistency and reducing token use in IEEE 118-bus tests.
-
AInstein: Can LLMs Solve Research Problems From Parametric Memory Alone?
LLMs generate valid solutions to over 70% of AI research problems from parametric memory alone but rediscover the exact published approach less than 19% of the time, with performance limited by cross-domain analogical transfer.
-
Retrieval-of-Thought: Efficient Reasoning via Reusing Thoughts
Retrieval-of-Thought organizes prior reasoning into a thought graph for retrieval and reward-guided recombination, reducing output tokens by up to 40% and latency by 82% while preserving accuracy on reasoning benchmarks.
-
PEER: Unified Process-Outcome Reinforcement Learning for Structured Empathetic Reasoning
PEER applies GRPO reinforcement learning with a unified process-outcome reward model to structured empathetic reasoning steps on the SER dataset, yielding gains in empathy, strategy alignment, and human-likeness.
-
GenoMAS: A Multi-Agent Framework for Scientific Discovery via Code-Driven Gene Expression Analysis
GenoMAS deploys six specialized LLM agents with guided planning to preprocess transcriptomic data and identify genes, reaching 89.13% composite similarity and 60.48% F1 on the GenoTEX benchmark while outperforming prior methods.
-
When LLM Rationales Become User-Facing: Effects on Trust Perception, Decision-Making, and Gaze Behaviors
Two linked user studies find that LLM rationale correctness and certainty framing affect trust and decision confidence while presentation format does not, and incorrect rationales increase gaze attention and pupil size.
-
ARIA: A Causal-Aware Framework for Rescuing LLM Reasoning in Trustworthy Materials Discovery
ARIA is a three-tier causal framework that conditions LLM knowledge use on mechanistic completeness for forward prediction and inverse design of 2D materials, producing auditable traces.
-
Characterize Then Distill: Mechanistic Reasoning in Large Output Spaces
Reasoning in large output spaces proceeds via shortlisting then fine-grained reasoning; this characterization enables a mechanistic distillation strategy that outperforms standard distillation.
-
Trust Region On-Policy Distillation
TrOPD stabilizes on-policy distillation for LLMs with trust-region learning, outlier estimation, and off-policy guidance, outperforming prior OPD methods on reasoning and code benchmarks.
-
EvoSci: A Bio-Inspired Multi-Agent Framework for the Evolution of Scientific Discovery
EvoSci combines evolutionary multi-agent collaboration with knowledge graphs to produce scientific ideas that score higher on LLM peer-review metrics than baselines.
-
Understanding Inference Scaling for LLMs: Bottlenecks, Trade-offs, and Performance Principles
Reasoning workloads shift LLM inference to a capacity-bound regime where KV-cache fragmentation limits data parallelism, tensor parallelism unlocks memory at the 32B scale, and MoE models require hybrid strategies to avoid routing latency.
-
Dynamic Skill Lifecycle Management for Agentic Reinforcement Learning
SLIM dynamically optimizes the active external skill set in agentic RL via leave-one-skill-out marginal contribution estimates and lifecycle operations, delivering a 7.1% average gain over baselines on ALFWorld and SearchQA while showing some skills remain externally useful.
-
How You Begin is How You Reason: Driving Exploration in RLVR via Prefix-Tuned Priors
IMAX trains soft prefixes with an InfoMax reward to drive diverse exploration in RLVR, yielding up to 11.60% gains in Pass@4 over standard RLVR across model scales.
-
How Well Do LLMs Perform on the Simplest Long-Chain Reasoning Tasks: An Empirical Study on the Equivalence Class Problem
Non-reasoning LLMs fail the equivalence class problem while reasoning LLMs perform better but remain incomplete, with difficulty peaking at phase transition for the former and maximum diameter for the latter.
-
UpstreamQA: A Modular Framework for Explicit Reasoning on Video Question Answering Tasks
UpstreamQA disentangles video reasoning by using LRMs for explicit upstream object identification and scene context before downstream LMM VideoQA, improving performance and interpretability on OpenEQA and NExTQA in some cases.
-
Efficient Test-Time Scaling via Temporal Reasoning Aggregation
TRACE aggregates answer consistency and confidence trajectory over multiple reasoning steps to decide when to halt inference, reducing token usage by 25-30% while keeping accuracy within 1-2% of full reasoning.
-
GroupRank: A Groupwise Paradigm for Effective and Efficient Passage Reranking with LLMs
GroupRank uses groupwise LLM reranking with answer-free data synthesis and a group-ranking reward to reach 65.2 NDCG@10 on BRIGHT while providing 6.4x faster inference than listwise baselines.
-
Deep Thinking by Markov Chain of Continuous Thoughts
MarCos modifies transformers to perform continuous multi-step reasoning by mapping thought-level continuous states directly to next-thought distributions, achieving substantial wall-clock speedups on math problems.
-
Explicit Reasoning Makes Better Judges: A Systematic Study on Accuracy, Efficiency, and Robustness
Thinking LLMs achieve ~10 percentage points higher accuracy than non-thinking ones on RewardBench with under 2x compute overhead, outperforming augmentation strategies that cost over 8x more while also showing better bias robustness.
-
Pruning Long Chain-of-Thought of Large Reasoning Models via Small-Scale Preference Optimization
LCPO reduces average LRM output length by over 50% across benchmarks via targeted preference optimization while preserving reasoning performance.
-
ReasonCache: Accelerating Large Reasoning Model Serving through KV Cache Sharing
ReasonCache reuses similar KV cache states across reasoning steps in LRMs via collaborative filtering to boost serving throughput by up to 89.2% while preserving accuracy.
-
Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models
A survey organizing techniques to achieve efficient reasoning in LLMs by shortening chain-of-thought outputs.
-
SuCo: Sufficiency-guided Continuous Adaptive Reasoning
SuCo defines minimal sufficient CoT and applies a two-stage fine-tuning plus RL framework to enable continuous adaptive reasoning control, claiming gains in both accuracy and token efficiency on math, code, and science tasks.
-
Evaluation of ML Resource Utilization Requires Model Life Cycle Assessment
The paper calls for life cycle assessment to capture embodied hardware costs and full pipeline operational costs in AI development and deployment.
-
Agentic Reasoning for Large Language Models
The survey structures agentic reasoning for LLMs into foundational, self-evolving, and collective multi-agent layers while distinguishing in-context orchestration from post-training optimization and reviewing applications across domains.
-
Large Language Models for Multi-Robot Systems: A Survey
A survey that categorizes LLM uses in multi-robot systems across task allocation, motion planning, action generation, and human interaction, while noting challenges and future research opportunities.
-
Reinforcement Learning for Scalable and Trustworthy Intelligent Systems
Reinforcement learning is advanced for communication-efficient federated optimization and for preference-aligned, contextually safe policies in large language models.
-
From System 1 to System 2: A Survey of Reasoning Large Language Models
The survey organizes the shift of LLMs toward deliberate System 2 reasoning, covering model construction techniques, performance on math and coding benchmarks, and future research directions.
-
Bridging Brains and Machines: A Unified Frontier in Neuroscience, Artificial Intelligence, and Neuromorphic Systems
A position and survey paper that identifies convergence between neuroscience, AGI, and neuromorphic computing and outlines four key integration challenges.
- The Periodic Table of LLM Reasoning: A Structured Survey of Reasoning Paradigms, Methods, and Failure Modes