| Step 3.5 Flash: Open Frontier-Level Intelligence with 11B Active Parameters | ✓ Link | 98.4 | Step 3.5 Flash (11B) | 2026-02-11 |
| Seed2.0 Model Card: Towards Intelligence Frontier for Real-World Complexity | | 97.3 | Seed2.0 Pro | 2026-06-30 |
| Kimi K2.5: Visual Agentic Intelligence | ✓ Link | 95.4 | Kimi K2.5 | 2026-02-02 |
| Claim-Level Reliability Assessment for Efficient Test-Time Reasoning | ✓ Link | 95.0 | Qwen3.5-27B + CLR@32 | 2026-08-12 |
| Unified Audio Intelligence Without Regressing on Text Intelligence | | 92.2 | Audex 30B-A3B | 2026-07-06 |
| STEP3-VL-10B Technical Report | ✓ Link | 92.14 | Step3-VL-10B (PaCoRe) | 2026-01-14 |
| Limite 1B - Violetto: Technical Report | ✓ Link | 91.35 | Limite 1B - Violetto | 2026-10-06 |
| MiMo-V2-Flash Technical Report | ✓ Link | 84.4 | MiMo-V2-Flash (Thinking) | 2026-01-06 |
| Thought-Level Beam Search for Reasoning | ✓ Link | 75.8 | Gambit (Phi-4-reasoning-plus-14B) | 2026-08-11 |
| PRISM: Pushing the Frontier of Deep Think via Process Reward Model-Guided Inference | ✓ Link | 75.6 | PRISM (gpt-oss-20b; LLM Aggregate) | 2026-03-03 |
| JudgeRLVR: Judge First, Generate Second for Efficient Reasoning | | 70.0 | JudgeRLVR (Qwen3-30B-A3B) | 2026-01-13 |
| Does On-Policy Distillation Really Distill? From Noisy Teacher to Self-Improvement | ✓ Link | 67.4 | OPSA (Qwen3.5-9B) | 2026-08-31 |
| Falcon-H1R: Pushing the Reasoning Frontiers with a Hybrid Model for Efficient Test-Time Scaling | ✓ Link | 64.9 | Falcon-H1R (7B) | 2026-01-05 |
| Verify Before You Distill: Prompt-Level Teacher Gating for On-Policy Distillation | | 64.27 | TGOPD (35B single-domain mask-only) | 2026-09-02 |
| Lightning Weave: Improving the Accuracy-Efficiency Frontier of Reasoning Models through Capability Composition | | 64 | Lightning Weave (Qwen3.5-4B) | 2026-09-13 |
| DoubtLess: Training-Free Contrastive Decoding for Efficient Reasoning | ✓ Link | 56.7 | DoubtLess (Qwen3-32B steered by Qwen3-0.6B) | 2026-10-13 |
| Anti-Self-Distillation for Reasoning RL via Pointwise Mutual Information | ✓ Link | 56.2 | Olmo3-7B-TK + AntiSD | 2026-05-12 |
| Is Next-Chunk Reasoning RL Really Better than SFT? Revisiting Training Strategies under no-CoT Data | | 55.0 | Mixed SFT + RLVR | 2026-08-24 |
| Value-Aware Stochastic KV Cache Eviction for Reasoning Models | ✓ Link | 50.0 | VaSE-AttnV (Qwen3-14B) | 2026-06-02 |
| On-Policy Self-Distillation without Any Supervision | ✓ Link | 50.0 | u-OPSD Qwen3-30B-A3B-Instruct-2507 | 2026-08-09 |
| KnowRL: Boosting LLM Reasoning via Reinforcement Learning with Minimal-Sufficient Knowledge Guidance | ✓ Link | 48.75 | KnowRL-Nemotron-1.5B (CSS) | 2026-04-14 |
| Better Supervision Is Nearby: Neighborhood On-Policy Self-Distillation | | 48.24 | N-OPSD (Qwen3-8B) | 2026-09-30 |
| Self-Distillation Zero: Self-Revision Turns Binary Rewards into Dense Supervision | ✓ Link | 45.4 | SD-Zero (Qwen3-4B-Instruct) | 2026-04-13 |
| Value-Aware Stochastic KV Cache Eviction for Reasoning Models | ✓ Link | 44.38 | VaSE-AttnV (Qwen3-4B) | 2026-06-02 |
| MOPD-Router: Rethinking Teacher Routing in Multi-Teacher On-Policy Distillation | ✓ Link | 40.83 | MOPD-Router (ExpertAlign, Qwen3-4B-Non-Thinking student, domain-labeled) | 2026-09-25 |
| JustRL: Scaling a 1.5B LLM with a Simple RL Recipe | ✓ Link | 40.63 | JustRL-Nemotron 1.5B | 2025-12-18 |
| Influence-Directed Distillation: Solving the Diversity Bottleneck in Sampled-Token On-Policy Distillation | | 37.6 | Qwen3-8B-Non-Thinking + IDA-OPD | 2026-08-30 |
| Weak-to-Strong On-Policy Distillation | ✓ Link | 35.1 | W2S-OPD (single-teacher contrast) | 2026-07-28 |
| Filter, Then Reweight: Rethinking Optimization Granularity in On-Policy Distillation | ✓ Link | 35.0 | FiRe-OPD | 2026-06-01 |
| FlowBalance: Verifier-Grounded Self-Improvement from On-Policy Reasoning Experience | ✓ Link | 34.67 | Qwen3-8B + FlowBalance | 2026-09-03 |
| Beyond Domain-Routed Supervision: Cross-Teacher Guarding for Multi-Teacher On-Policy Distillation | | 33.54 | Road MOPD (Qwen3-4B) | 2026-10-09 |
| Eliciting Weak-to-Strong Generalization with On-Policy Reverse Distillation | ✓ Link | 31.42 | Qwen3-8B + OPRD | 2026-09-08 |
| SAF-OPD: Stable Advantage Fusion for On-Policy Distillation | | 31.15 | SAF (Qwen3-4B) | 2026-07-31 |
| MHPO: Modulated Hazard-aware Policy Optimization for Stable Reinforcement Learning | | 17.9 | MHPO (Qwen2.5-Math-7B-Instruct) | 2026-03-14 |