探索利用论文趋势 2023–2026

主要结论

• 探索利用论文从 2023 年的 47 篇增长到 2026 年的 71 篇，三年间扩大到约 1.5 倍。
• NeurIPS 是发探索利用论文最多的会议，2023–2026 年共 93 篇，占四个会议总量的 40%。
• 交叉方向里涨得最快的是推理模型：同时属于两个主题的论文从 2023 年的 1 篇增至 2026 年的 10 篇。
• 在「推荐与广告」的 5 个主题里，它的 2026 年论文数排第 4；同组增长最快的点击率预测是 2.6 倍，探索利用是 1.5 倍。

代表论文

按年份和会议均匀挑选，方便快速看到每一届的样貌

• NeurIPS 2026Decentralized Q-Learning in Markov Potential Games
• ICML 2026Bandit Social Leaning Dynamics with Exploration Episodes
• ICLR 2026Laplacian Kernelized Bandit
• CVPR 2026Underground Plant Exploration Non-Destructive 3D Root Assessment with GPR Based
• NeurIPS 2025DUET: Dual-Perspective Pseudo Labeling and Uncertainty-aware Exploration & Exploitation Training for Source-Free Domain Adaptation
• ICML 2025Clustering Items through Bandit Feedback: Finding the Right Feature out of Many
• ICLR 2025B-STaR: Monitoring and Balancing Exploration and Exploitation in Self-Taught Reasoners
• CVPR 2025Dynamic Pseudo Labeling via Gradient Cutting for High-Low Entropy Exploration
• NeurIPS 2024Exploration, Exploitation, and Engagement in Multi-Armed Bandits with Abandonment
• ICML 2024A Contextual Combinatorial Bandit Approach to Negotiation
• ICLR 2024Dropout-Based Rashomon Set Exploration for Efficient Predictive Multiplicity Estimation
• CVPR 2024Instance-aware Exploration-Verification-Exploitation for Instance ImageGoal Navigation
• NeurIPS 2023Adaptive Algorithms for Relaxed Pareto Set Identification
• ICML 2023Curiosity in Hindsight: Intrinsic Exploration in Stochastic Environments
• ICLR 2023Combinatorial Pure Exploration of Causal Bandits
• CVPR 2023Exact-NeRF: An Exploration of a Precise Volumetric Parameterization for Neural Radiance Fields
• NeurIPS 2026Hit Expansion via Localized Exploration of Synthesizable Chemical Space
• ICML 2026Exploration-free Algorithms for Multi-group Mean Estimation
• ICLR 2026DIVERSE: Disagreement-Inducing Vector Evolution for Rashomon Set Exploration
• CVPR 2026From Exploration to Exploitation A Two-Stage Entropy RLVR Approach for

全部探索利用论文

共 71 篇，第 1 / 2 页

1. CVPR 2026DreamSAC Learning Hamiltonian World Models via Symmetry Exploration
2. CVPR 2026From Exploration to Exploitation A Two-Stage Entropy RLVR Approach for
3. CVPR 2026ReLaX Reasoning with Latent Exploration for Large Reasoning Models
4. CVPR 2026Underground Plant Exploration Non-Destructive 3D Root Assessment with GPR Based
5. ICLR 2026Beyond Noisy-TVs: Noise-Robust Exploration Via Learning Progress Monitoring
6. ICLR 2026Contextual Similarity Distillation: Ensemble Uncertainties with a Single Model
7. ICLR 2026Controllable Exploration in Hybrid-Policy RLVR for Multi-Modal Reasoning
8. ICLR 2026Demystifying Emergent Exploration in Goal-Conditioned RL
9. ICLR 2026Discovering Diverse Behaviors via Temporal Contrastive Learning
10. ICLR 2026DIVERSE: Disagreement-Inducing Vector Evolution for Rashomon Set Exploration
11. ICLR 2026EEPO: Exploration-Enhanced Policy Optimization via Sample-Then-Forget
12. ICLR 2026Exploration vs Exploitation: Rethinking RLVR through Clipping, Entropy, and Spurious Reward
13. ICLR 2026Flow Expansion via Verifier-Constrained Noised State Space Exploration
14. ICLR 2026In-Context Learning for Pure Exploration
15. ICLR 2026Instance-Dependent Fixed-Budget Pure Exploration in Reinforcement Learning
16. ICLR 2026Is Pure Exploitation Sufficient in Exogenous MDPs with Linear Function Approximation?
17. ICLR 2026Laplacian Kernelized Bandit
18. ICLR 2026Learn the Ropes, Then Trust the Wins: Self-imitation with Progressive Exploration for Agentic Reinforcement Learning
19. ICLR 2026Multi-Armed Bandits with Minimum Aggregated Revenue Constraints
20. ICLR 2026Next-ToBE: Probabilistic Next Token-Bag Exploitation for Activating Anticipatory Capacity in LLMs
21. ICLR 2026Non-Asymptotic Analysis of (Sticky) Track-and-Stop
22. ICLR 2026Risk-Sensitive Reinforcement Learning for Alleviating Exploration Dilemmas in Large Language Models
23. ICLR 2026Sample-efficient and Scalable Exploration in Continuous-Time RL
24. ICLR 2026Strategic Scaling of Test-Time Compute: A Bandit Learning Approach
25. ICLR 2026Token Hidden Reward: Steering Exploration-Exploitation in Group Relative Deep Reinforcement Learning
26. ICLR 2026Universal Value-Function Uncertainties
27. ICLR 2026When Greedy Wins: Emergent Exploitation Bias in Meta-Bandit LLM Training
28. ICML 2026Bandit Social Leaning Dynamics with Exploration Episodes
29. ICML 2026Bilinear Bandits with Partially Observable Features
30. ICML 2026Covariance Volume Maximization for Embodied Latent Exploration in Deep Reinforcement Learning
31. ICML 2026Creat3r: Confidence Reaggregation for Exploration-aware Active 3D Reconstruction
32. ICML 2026CUPID in the Model Zoo: Online Matchmaking for Selecting Your Dream LLM
33. ICML 2026Data-Source Adaptive Online Learning under Heteroscedastic Noise
34. ICML 2026DecFus: Decentralized Layer-wise Fusion with Dynamic Exploration and Exploitation
35. ICML 2026DF-ExpEnse: Diffusion Filtered Exploration for Sample Efficient Finetuning
36. ICML 2026Exploration-free Algorithms for Multi-group Mean Estimation
37. ICML 2026Follow-the-Perturbed-Leader for Decoupled Bandits: Best-of-Both-Worlds and Practicality
38. ICML 2026From Bits to Rounds: Parallel Decoding with Exploration for Diffusion Language Models
39. ICML 2026HyPER: Bridging Exploration and Exploitation for Scalable LLM Reasoning with Hypothesis Path Expansion and Reduction
40. ICML 2026Improved Bounds for Reward-Agnostic and Reward-Free Exploration
41. ICML 2026Incentivized Exploration with Stochastic Covariates: A Two-Stage Mechanism Design for Recommender System
42. ICML 2026Long Live The Balance: Information Bottleneck Driven Tree-based Policy Optimization
43. ICML 2026OneSearch: A Preliminary Exploration of the Unified End-to-End Generative Framework for E-commerce Search
44. ICML 2026Optimal Top-$k$ Identification from Pairwise Comparisons
45. ICML 2026Tracking Drift: Variation-Aware Entropy Scheduling for Non-Stationary Reinforcement Learning
46. ICML 2026XRPO: Pushing the Limits of GRPO with Targeted Exploration and Exploitation
47. NeurIPS 2026Agnostic Online Learning with Reliable Abstention
48. NeurIPS 2026Boosting Inference with Guided Reasoning: Stochastic Exploration for Recursive Models
49. NeurIPS 2026CIG: Exploration via Conditional Information Gain
50. NeurIPS 2026Contrastive Pretraining Scales Agentic Exploration

常见问题

2023–2026 年四大会议一共有多少篇探索利用论文？

共 235 篇，按年份分别是 2023 年 47 篇、2024 年 63 篇、2025 年 54 篇、2026 年 71 篇。

NeurIPS 2026 有多少篇探索利用论文？

NeurIPS 2026 有 25 篇探索利用论文，上一届是 20 篇。

ICLR 2026 有多少篇探索利用论文？

ICLR 2026 有 23 篇探索利用论文，上一届是 10 篇。

ICML 2026 有多少篇探索利用论文？

ICML 2026 有 19 篇探索利用论文，上一届是 21 篇。

CVPR 2026 有多少篇探索利用论文？

CVPR 2026 有 4 篇探索利用论文，上一届是 3 篇。

哪个会议的探索利用论文最多？

NeurIPS，2023–2026 年共 93 篇。