OpenCodePapers

multiple-choice-question-answering-mcqa-on-gpqa-diamond

Question AnsweringMultiple Choice Question Answering (MCQA)
Dataset Link
Results over time
Click legend items to toggle metrics. Hover points for model names.
Leaderboard
Click a sortable column header to sort.
PaperCodeAccuracyModelNameReleaseDate
GPT-6 Astra: A new generation of intelligence✓ Link96.0GPT-6 Astra2026-09-03
GPT-5.6: Frontier intelligence that scales with your ambition✓ Link94.6GPT-5.6 Sol2026-07-09
Introducing GPT-5.494.4GPT-5.4 Pro2026-03-05
Gemini 3.1 Pro: A smarter model for your most complex tasks94.3Gemini 3.1 Pro2026-02-19
Introducing Claude Opus 4.794.2Claude Opus 4.72026-04-16
Claude Fable 5 and Claude Mythos 5✓ Link94.1Claude Mythos 52026-06-09
A new era of intelligence with Gemini 393.8Gemini 3 Deep Think2025-11-18
Introducing GPT-5.593.6GPT-5.52026-04-23
Introducing Claude Opus 4.893.6Claude Opus 4.82026-05-28
Kimi K3: Open Frontier Intelligence✓ Link93.5Kimi K3 (max)2026-07-27
Introducing GPT-5.293.2GPT-5.2 Pro2025-12-11
Qwen3.8-Max: A New Bar for Coding and Cowork✓ Link92.6Qwen3.8-Max2026-08-03
Hy4 preview✓ Link92.3Hy4 preview2026-08-28
On the Design of Qwen3.8-Next Architecture: Evaluation, Efficiency, and Training Stability✓ Link91.7Qwen3.8-Flash-Next2026-08-31
Introducing Claude Opus 4.691.3Claude Opus 4.62026-02-05
GLM-5.2: Built for Long-Horizon Tasks✓ Link91.2GLM-5.22026-06-16
DeepSeek-V4.1-Flash: Pushing the Limits of KV Cache Compression✓ Link90.9DeepSeek-V4.1-Flash Max2026-09-17
Kimi K2.6✓ Link90.5Kimi K2.62026-04-20
Introducing Beam: Reflection's 501B open-weight model90.5Beam2026-10-05
Introducing Gemini 3 Flash90.4Google Gemini 3 Flash (Thinking)2025-12-17
Hy3✓ Link90.4Hy32026-07-02
DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence✓ Link90.1DeepSeek-V4-Pro Max2026-04-26
Darwin Family: MRI-Trust-Weighted Evolutionary Merging for Training-Free Scaling of Language-Model Reasoning90.0Darwin-35B-A3B-Opus2026-05-14
Introducing Claude Sonnet 4.689.9Claude Sonnet 4.62026-02-17
The MiniMax-M2 Series: Mini Activations Unleashing Max Real-World Intelligence✓ Link89.8MiniMax-M2.72026-05-26
Inkling-Small✓ Link89.5Inkling-Small2026-07-27
Qwen3.8-27B✓ Link89.2Qwen3.8-27B2026-08-05
Ornith-1.5: From Self-Scaffolding to Self-Improvement89.2Ornith-1.5-35B-A3B2026-08-20
Introducing Muse Spark Open, The Most Capable US Open-Weight Model to Date✓ Link89.0Muse Spark Open2026-10-15
Qwen3.5: Towards Native Multimodal Agents✓ Link88.4Qwen3.5-397B-A17B2026-02-16
DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence✓ Link88.1DeepSeek-V4-Flash Max2026-04-26
Introducing LongCat-2.0✓ Link87.9LongCat-2.02026-07-05
Qwen3.6✓ Link87.8Qwen3.6-27B2026-04-21
Kimi K2.5: Visual Agentic Intelligence✓ Link87.6Kimi K2.52026-02-02
Nanbeige4.2-3B: Unlocking Agentic Capabilities in a Compact Model✓ Link87.4Nanbeige4.2-3B2026-07-21
InternAgent-1.5: A Unified Agentic Framework for Long-Horizon Autonomous Scientific Discovery✓ Link87.37InternAgent-1.52026-02-09
Streaming Communication in Multi-Agent Reasoning✓ Link87.37STREAM-MA (Claude Opus 4.6, Chain)2026-06-03
Magenta: Closing the Loop Between Mathematical Reasoning and Lean Verification87.3K2-Horizon-375B-A23B2026-09-10
Hy3 preview✓ Link87.2Hy3 preview2026-04-23
Introducing Claude Opus 4.587.0Claude Opus 4.52025-11-24
Nemotron 3 Ultra: Open, Efficient Mixture-of-Experts Hybrid Mamba-Transformer Model for Agentic Reasoning✓ Link87.0Nemotron 3 Ultra2026-06-03
Qwen3.5: Towards Native Multimodal Agents✓ Link86.6Qwen3.5-122B-A10B2026-02-16
Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities86.4Gemini 2.5 Pro2025-07-07
ERNIE 5.0 Technical Report86.36ERNIE 5.02026-02-04
GLM-5.1: Towards Long-Horizon Tasks✓ Link86.2GLM-5.12026-04-03
GLM-5: From Vibe Coding to Agentic Engineering✓ Link86.0GLM-52026-02-11
Qwen3.6✓ Link86.0Qwen3.6-35B-A3B2026-04-21
A*-Thought-V2: Efficient Latent Reasoning via Geometric Dynamics of LLM✓ Link86.0A*-Thought-V2 (Qwen3.6-27B, tau=90)2026-09-07
Ling and Ring 2.6 Technical Report: Efficient and Instant Agentic Intelligence at Trillion-Parameter Scale✓ Link85.89Ring-2.6-1T (xhigh)2026-06-13
GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models✓ Link85.7GLM-4.72025-08-08
DeepSeek-V3.2 Technical Report✓ Link85.7DeepSeek-V3.2-Speciale2025-12-02
Qwen3.5: Towards Native Multimodal Agents✓ Link85.5Qwen3.5-27B2026-02-16
ParaTempo: Efficient Parallel Reasoning via Temporal Confidence✓ Link85.4ParaTempo (Qwen3.5-35B-A3B)2026-08-17
Thomson: Continual Learning of Frontier Models for SovereignAI85.4Thomson-1.0-Small2026-08-24
MiniMax-M2.5✓ Link85.2MiniMax-M2.52026-02-12
Step 3.5 Flash: Open Frontier-Level Intelligence with 11B Active Parameters✓ Link85.0Step-3.5-Flash PaCoRe2026-02-11
AgentMercury: Your Agent Can Synthesize Verifiable Environments for Business Scenarios at scale85.0Qwen3.5-35B-A3B + GRPO + Ours2026-08-21
MiMo-V2-Flash Technical Report✓ Link84.3MiMo-V2-Flash (Thinking)2026-01-06
Gemma 4 Technical Report✓ Link84.3Gemma 4 31B2026-07-02
Kolibri: A Sovereign European Model on the Pareto Frontier✓ Link84.3Kolibri2026-10-03
Qwen3.5: Towards Native Multimodal Agents✓ Link84.2Qwen3.5-35B-A3B2026-02-16
MAI-Thinking-1: Building a Hill-Climbing Machine84.2MAI-Thinking-12026-06-02
Introducing efficiency focused "Thinking CAP" model series✓ Link83.8ThinkingCap-Qwen3.6-27B2026-07-06
Step 3.5 Flash: Open Frontier-Level Intelligence with 11B Active Parameters✓ Link83.5Step-3.5-Flash2026-02-11
Muse Glimmer 30B: An On-Device Agentic Model✓ Link83.5Muse Glimmer-30B (High Reasoning)2026-10-17
Introducing Claude Sonnet 4.583.4Claude Sonnet 4.52025-09-29
Motif 3: Technical Report✓ Link83.4Motif 32026-08-10
Introducing OpenAI o3 and o4-mini✓ Link83.3OpenAI o32025-04-16
Mach-Mind-4-Flash Technical Report83.08Mach-Mind-4-Flash2026-07-10
Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities82.8Gemini 2.5 Flash2025-07-07
Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding✓ Link82.8Qwen3.5-35B-A3B + Confident Decoding2026-06-20
Gemma 4 Technical Report✓ Link82.3Gemma 4 26B-A4B2026-07-02
Nanbeige4-3B Technical Report: Exploring the Frontier of Small Language Models✓ Link82.2Nanbeige4-3B-Thinking-25112025-12-06
Qwen3.5: Towards Native Multimodal Agents✓ Link81.7Qwen3.5-9B2026-02-16
Optical Reasoning: Rethinking Images as an Expressive Reasoning Medium Beyond Text✓ Link81.31T-OR (Claude Sonnet 4.5, 259.3 tokens)2026-06-08
SecOPD: Mitigating Adaptive Prompt Injections by On-Policy Distillation✓ Link81.3SecOPD (Qwen3.6-27B)2026-08-21
Claude Opus 4.180.9Claude Opus 4.12025-08-05
Intern-S2-Mobius: Foundation Model with Decoupled Knowledge and Reasoning✓ Link80.81Intern-S2-Mobius-35B2026-08-14
EXAONE 4.5 Technical Report✓ Link80.5EXAONE 4.5 (Reasoning)2026-04-09
The Other Half of the Memory Wall: Serving 35B MoEs from SSD with Trained Routing Prediction✓ Link79.8Edge0-35B-A3B-preview2026-09-16
Scalable Visual Pretraining for Language Intelligence79.29Qwen3.5-35B-A3B + VP2026-07-10
K-EXAONE Technical Report✓ Link79.1K-EXAONE (236B, reasoning)2026-01-05
Gemma 4 Technical Report✓ Link78.8Gemma 4 12B Unified2026-07-02
IndexCache: Accelerating Sparse Attention via Cross-Layer Index Reuse✓ Link78.6DSA (30B) + IndexCache (1/4, searched pattern)2026-03-12
OmniEdu: Open Foundation Models for Learning and Teaching✓ Link77.78OmniEdu-27B2026-09-19
Introducing OpenAI o177.3OpenAI o12024-09-12
Efficient Agentic Reasoning Through Self-Regulated Simulative Planning✓ Link77.3SR 2 AM-v1.0-30B2026-05-21
Thought-Level Beam Search for Reasoning✓ Link77.1Gambit (Phi-4-reasoning-plus-14B)2026-08-11
Unlocking Lossless Speedups in LLMs via Discrete Diffusion✓ Link77.1Uno (K2-Horizon-7B-Uno)2026-09-03
QwenLong-L1.5: Post-Training Recipe for Long-Context Reasoning and Memory Management✓ Link76.78QwenLong-L1.5-30B-A3B2025-12-15
Ling and Ring 2.6 Technical Report: Efficient and Instant Agentic Intelligence at Trillion-Parameter Scale✓ Link76.17Ling-2.6-1T2026-06-13
Nemotron-Cascade 2: Post-Training LLMs with Cascade RL and Multi-Domain On-Policy Distillation76.1Nemotron-Cascade-2-30B-A3B2026-03-19
DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning✓ Link75.8DeepSeek-R1-Zero2025-01-22
Stop When Reasoning Converges: Semantic-Preserving Early Exit for Reasoning Models✓ Link75.8Qwen3-30B-A3B-Thinking with PUMA2026-05-17
SPADE: Self-Play in Adaptive Synthetic Executable Environments✓ Link75.8SPADE (Games, Qwen3-30B-A3B-Instruct-2507)2026-08-19
Rufus-Air: An Open LLM Post-Training Recipe75.6Rufus-Air2026-09-24
NVIDIA Nemotron 3.5 Lightning✓ Link75.44NVIDIA Nemotron 3.5 Lightning 30B-A3B2026-08-11
EXAONE 4.0: Unified Large Language Models Integrating Non-reasoning and Reasoning Modes✓ Link75.4EXAONE 4.0 32B (Reasoning)2025-07-15
Kimi K2: Open Agentic Intelligence✓ Link75.1Kimi K22025-07-28
Introducing Claude 474.9Claude Opus 42025-05-22
Unified Audio Intelligence Without Regressing on Text Intelligence74.9Audex 30B-A3B2026-07-06
LFM2.5-24B-A2B: Built for Fast On-Device Agents74.42LFM2.5-24B-A2B2026-10-13
INTELLECT-3: Technical Report✓ Link74.4INTELLECT-32025-12-18
STEP3-VL-10B Technical Report✓ Link73.99Step3-VL-10B (PaCoRe)2026-01-14
Does On-Policy Distillation Really Distill? From Noisy Teacher to Self-Improvement✓ Link73.7OPSA (Qwen3.5-9B)2026-08-31
LongCat-Flash Technical Report✓ Link73.23LongCat-Flash2025-09-01
DiffusionGemma Technical Report✓ Link73.2DiffusionGemma 26B A4B2026-07-31
Believe Your Model: Distribution-Guided Confidence Calibration✓ Link73.18Qwen3-32B with DistriVoting (DIS-GMM*)2026-03-04
Qwen3-Omni Technical Report✓ Link73.1Qwen3-Omni-Flash-Thinking2025-09-22
Qwen3-Next72.9Qwen3-Next-80B-A3B-Instruct2025-09-09
VibeThinker-3B: Exploring the Frontier of Verifiable Reasoning in Small Language Models✓ Link72.9VibeThinker-3B (CLR)2026-06-15
One Model to Loop Them All: Elastic Latent Recurrence for Language Models72.8Koi-4B2026-10-08
Closing the Loop: Practical Training Recipes for Looped Language Models72.41Dory1.0-4B2026-09-30
Nemotron 3 Nano Omni: Efficient and Open Multimodal Intelligence✓ Link72.2Nemotron 3 Nano Omni2026-04-27
PRISM: Pushing the Frontier of Deep Think via Process Reward Model-Guided Inference✓ Link71.7PRISM (gpt-oss-20b; Majority Vote)2026-03-03
Qwen3 Technical Report✓ Link71.1Qwen3-235B-A22B (Thinking)2025-05-14
ZAYA1-8B Technical Report✓ Link71.0ZAYA1-8B2026-05-06
ACC: Compiling Agent Trajectories for Long-Context Training70.2Qwen3-30B-A3B-Thinking + ACC2026-05-21
MiniCPM5-2B✓ Link70.2MiniCPM5-2B2026-09-07
CHIMERA: Compact Synthetic Data for Generalizable LLM Reasoning70.1Qwen3-4B-Thinking-2507 (4B) fine-tuned on Chimera (SFT+RL)2026-03-01
Introducing Claude 470.0Claude Sonnet 42025-05-22
The Llama 4 herd: The beginning of a new era of natively multimodal AI innovation✓ Link69.8Llama 4 Maverick2025-04-05
Nemotron-Cascade: Scaling Cascaded Reinforcement Learning for General-Purpose Reasoning Models69.6Nemotron-Cascade (14B, Thinking)2025-12-15
Solar Open Technical Report✓ Link68.1Solar Open (102B)2026-01-11
Claude 3.7 Sonnet and Claude Code68.0Claude 3.7 Sonnet2025-02-24
Quantization-Aware Healing: A Practical Recipe for Recovering Compressed, 4-Bit LLMs67.4Hypernova-60B2026-08-21
Efficient Reasoning with Balanced Thinking✓ Link67.2ReBalance (QwQ-32B)2026-03-12
JudgeRLVR: Judge First, Generate Second for Efficient Reasoning66.9Judge Only (Qwen3-30B-A3B)2026-01-13
Scaling Embeddings Outperforms Scaling Experts in Language Models66.78LongCat-Flash-Lite (68.5B)2026-01-29
MiMo-V2.5-Pro✓ Link66.7MiMo-V2.5-Pro2026-04-27
Recursive Multi-Agent Systems✓ Link66.2RecursiveMAS2026-04-28
Reasoning over mathematical objects: on-policy reward modeling and test time aggregation63.83Principia-4B2026-03-19
Arcee Trinity Large Technical Report✓ Link63.32Trinity Large Preview2026-02-19
Rethinking Generalization in Reasoning SFT: A Conditional Analysis on Optimization, Data, and Model Capability✓ Link63.3Qwen3-14B (Math-CoT SFT)2026-04-08
Qwen3 Technical Report✓ Link62.9Qwen3-235B-A22B (Non-thinking)2025-05-14
Introducing Gemini 2.0: our new AI model for the agentic era62.1Gemini 2.0 Flash Experimental2024-12-11
Introspective Diffusion Language Models✓ Link62.1I-DLM (32B)2026-04-13
CopT: Contrastive On-Policy Thinking with Continuous Spaces for General and Agentic Reasoning✓ Link61.62CopT (Qwen3-8B)2026-05-19
Is Next-Chunk Reasoning RL Really Better than SFT? Revisiting Training Strategies under no-CoT Data60.98Mixed SFT + RLVR2026-08-24
Recursive Think-Answer Process for LLMs and VLMs60.7MiMo-RL-7B-R-TAP2026-03-02
Zone of Proximal Policy Optimization: Teacher in Prompts, Not Gradients60.6ZPPO (Qwen3.5-2B)2026-06-16
Hunyuan Dense V1✓ Link60.1Hunyuan-7B-Instruct2025-07-30
LSRIF: Logic-Structured Reinforcement Learning for Instruction Following60.1Distill-Qwen-14B (LsrIF)2026-01-10
Introducing Claude 3.5 Sonnet59.4Claude 3.5 Sonnet2024-06-20
DeepSeek-V3 Technical Report✓ Link59.1DeepSeek-V32024-12-27
RISE: Recursive Improvement via Self-Extrapolating Policy Distillation59.0RISE (weight,Qwen3-8B)2026-09-04
Agentic Chain-of-Thought Steering for Efficient and Controllable LLM Reasoning✓ Link58.6ACTS (Qwen3-8B)2026-06-02
Gemma 4 Technical Report✓ Link58.6Gemma 4 E4B2026-07-02
Self-Hinting Language Models Enhance Reinforcement Learning✓ Link57.8SAGE (Qwen3-4B-Instruct)2026-02-03
Mellum2 Technical Report✓ Link57.6Mellum 2 Thinking2026-05-29
Value-Aware Stochastic KV Cache Eviction for Reasoning Models✓ Link57.39VaSE-AttnV (Qwen3-14B)2026-06-02
The Llama 4 herd: The beginning of a new era of natively multimodal AI innovation✓ Link57.2Llama 4 Scout2025-04-05
DelTA: Discriminative Token Credit Assignment for Reinforcement Learning from Verifiable Rewards✓ Link56.67DelTA (Qwen3-14B-Base)2026-05-20
Weak-to-Strong On-Policy Distillation✓ Link56.5W2S-OPD (pre-/post-RL contrast)2026-07-28
Rubric-based On-policy Distillation✓ Link55.05ROPD (Qwen3-4B)2026-05-08
Composition-RL: Compose Your Verifiable Prompts for Reinforcement Learning of Large Language Models✓ Link54.7Qwen3-14B-Base + Composition-RL (12K subset)2026-02-12
MiniMax-01: Scaling Foundation Models with Lightning Attention✓ Link54.4MiniMax-Text-012025-01-14
SFT Conflicts, RL Coexists: A Theoretical and Empirical Analysis of Multi-Task Learning for LLMs✓ Link54.0Adapted Parallel-RL (DeepSeek-R1-Distill-Qwen-7B)2026-08-06
From P(y|x) to P(y): Investigating Reinforcement Learning in Pre-train Space✓ Link53.48DSRL (Qwen3-8B)2026-04-15
Agentic Critical Training53.37ACT (Qwen3-8B)2026-03-09
PRISM: Demystifying Retention and Interaction in Mid-Training52.86Granite-3.3 (8B) PRISM (MCS) + RL (MC)2026-03-17
dots.llm1 Technical Report✓ Link52.6dots.llm1.inst2025-06-06
NGM: A Plug-and-Play Training-Free Memory Module for LLMs✓ Link52.02Qwen3-14B + NGM2026-05-16
Phi-4-Mini-Reasoning: Exploring the Limits of Small Reasoning Language Models in Math52.0Phi-4-Mini-Reasoning2025-04-30
Introducing the next generation of Claude50.4Claude 3 Opus2024-03-04
Falcon-H1: A Family of Hybrid-Head Language Models Redefining Efficiency and Performance✓ Link49.66Falcon-H1-34B-Instruct2025-07-30
Surgical Post-Training: Cutting Errors, Keeping Knowledge✓ Link49.5SPoT-BCE (Qwen3-8B)2026-03-02
δ-mem: Efficient Online Memory for Large Language Models✓ Link49.49δ-mem (MSW) (Qwen3-8B)2026-05-12
Phi-4-Mini Technical Report: Compact yet Powerful Multimodal Language Models via Mixture-of-LoRAs✓ Link49.0Phi-4-Mini reasoning2025-02-24
InftyThink+: Effective and Efficient Infinite-Horizon Reasoning via Reinforcement Learning✓ Link48.99InftyThink+ (Qwen3-4B-Base, task-only RL)2026-02-06
Youtu-LLM: Unlocking the Native Agentic Potential for Lightweight Large Language Models✓ Link48.0Youtu-LLM 2B2025-12-31
DiagEvo: Diagnosis-Guided Self-Evolution via Hierarchical Error Memory47.5DiagEvo (Qwen3-8B, 235B-A22B diagnostician)2026-09-01
ProFit: Leveraging High-Value Signals in SFT via Probability-Guided Token Selection✓ Link46.53ProFit (Qwen3-14B-Base)2026-01-14
DARC: Decoupled Asymmetric Reasoning Curriculum for LLM Evolution✓ Link44.4DARC (Qwen3-8B-Base)2026-01-20
Attention Residuals✓ Link44.4Kimi Linear (48B) with Block AttnRes2026-03-16
A Sovereign, Open-Source Foundation Model for German and English✓ Link43.4Soofi S 30B-A3B Base2026-07-10
Semifactual Credit-Augmented Policy Optimization✓ Link42.45SCAPO (Qwen3-4B-Base)2026-09-30
SmolLM3: smol, multilingual, long-context reasoner✓ Link41.7SmolLM3 3B (Extended Thinking)2025-07-08
Introducing the next generation of Claude40.4Claude 3 Sonnet2024-03-04
Encoder-Decoder Gemma: Improving the Quality-Efficiency Trade-Off via Adaptation✓ Link40.4T5Gemma 9B-9B PrefixLM IT2025-04-08
Learning to Hint for Reinforcement Learning✓ Link40.4HiLL (Qwen2.5-7B-Instruct)2026-04-01
Geometry Conflict: Explaining and Controlling Forgetting in LLM Continual Post-Training✓ Link39.9Qwen3-14B with GCWM2026-05-10
Youtu-VL: Unleashing Visual Potential via Unified Vision-Language Supervision✓ Link39.8Youtu-VL (4B)2026-01-27
Difficulty-Adaptive Tree-Structured Policy Optimization for Expanding Reasoning Coverage in RLVR✓ Link39.6DATPO (Qwen3-4B-Base)2026-09-08
LFM-1B-Math: Can Small Models Be Concise Reasoners?39.29LFM-1.3B-Distill (32K output budget)2025-06-24
Small Language Models as Judges for Rubric-Based Reinforcement Learning38.8Probe-reward policy2026-08-30
Co-RL: Unsupervised Reasoning Emerges from Diverse Cohort in Multi-agent RL✓ Link37.9Co-RL (Qwen2.5-7B, Different family+)2026-08-19
LFM2.5-1.2B-Thinking: On-Device Reasoning Under 1GB37.86LFM2.5-1.2B-Thinking2026-01-20
Trust Region On-Policy Distillation36.24TrOPD (DeepSeek-R1-Distill-Qwen-1.5B)2026-05-31
Data-Efficient Autoregressive-to-Diffusion Language Models via On-Policy Distillation✓ Link36.1OPDLM-8B2026-06-04
MobileMoE: Scaling On-Device Mixture of Experts33.8MobileMoE-L (922M)2026-05-26
Introducing the next generation of Claude33.3Claude 3 Haiku2024-03-04
Queryable LoRA: Instruction-Regularized Routing Over Shared Low-Rank Update Atoms32.3000Instruction-Queryable LoRA2026-05-08
LFM2.5-350M: No Size Left Behind30.64LFM2.5-350M2026-03-31
Understanding Evolution Strategies for LLM Reasoning: Broader Reasoning Coverage than GRPO30.6DeepSeek-R1-Distill-Qwen-1.5B (GRPO to ES)2026-08-27
SPOT: Sparse Probing and Outcome Calibration for On-Policy Distillation✓ Link29.42SPOT Qwen3-1.7B-Base2026-08-05
LFM2 Technical Report✓ Link29.29LFM2-8B-A1B (0-shot)2025-11-28
KL for a KL: On-Policy Distillation with Control Variate Baseline28.6Qwen3-1.7B → Qwen3-1.7B-Base with v OPD (top-k)2026-05-08
T5Gemma 2: Seeing, Reading, and Understanding Longer27.8T5Gemma 2 4B-4B (post-trained)2025-12-16
MiniCPM5-1B✓ Link26.26MiniCPM5-1B (Thinking)2026-05-25
LFM2.5-230M: Built to Run Anywhere25.41LFM2.5-230M2026-06-25
Check out how to contribute new results. Then edit this file on GitLab or GitHub.