2026-05
- May 30, 2026 Agents Start Improving Themselves, and Reaching for Fewer Tools Daily
- May 29, 2026 Vision Models Start Redesigning How They Output Daily
- May 28, 2026 Diffusion Swallows the Decoder Too Daily
- May 27, 2026 The Rulers We Use to Measure What Models Really Think Are Broken Daily
- May 25, 2026 Agent Trajectories Let a 30B Match a 235B Daily
- May 24, 2026 Gated DeltaNet-2 Splits the Gate, Maestro Outscores GPT-5 Daily
- May 23, 2026 Optimizer Choice Stretches Capacity Scaling 2.3x Daily
- May 22, 2026 $15 Per Paper, Healthcare Agents Cap at 28% Daily
- May 21, 2026 Dual-Stream MoE Unifies Multimodal, Garment Video 30x Faster Daily
- May 20, 2026 Stop When Reasoning Converges, Save 26% of Tokens Daily
- May 19, 2026 8% of Tokens Decide the Reasoning Gap Daily
- May 18, 2026 Real-Time Video's Bottleneck Moved Past Step Count Daily
- May 17, 2026 Olympiad Gold Becomes a Two-Step Recipe Daily
- May 16, 2026 Readable Rules Don't Belong in LLM Weights Daily
- May 15, 2026 δ-mem Trades Long Context for an 8×8 State Matrix Daily
- May 14, 2026 Flow-OPD Lifts GenEval From 63 to 92 Daily
- May 13, 2026 Geometry Conflict Predicts Continual Fine-Tuning Forgetting Daily
- May 12, 2026 Soohak Caps Top Models at 30% Daily
- May 11, 2026 Lorem Ipsum Rescues GRPO's Wasted Hard Samples Daily
- May 9, 2026 10.6k SFT Trajectories Match Full RL Pipeline; Mamba Beats LZMA Daily
- May 8, 2026 T²PO Stabilizes Multi-Turn RL; MotionCache Cuts Video Steps 6x Daily
- May 7, 2026 Gradient Boosting Turns Out to Be Diffusion's Asymptotic Optimum Daily
- May 4, 2026 ViT Pre-Trains Like an LLM, Skips the CLIP Stage Daily
- May 3, 2026 FD as Loss: One-Step Generation Hits 0.72 FID Daily
- May 2, 2026 Cross-Architecture Distillation Shrinks dLLMs to 0.6B Daily
- May 1, 2026 Recursive MAS Cuts Tokens 35%, T2I Repaints Instead of Editing Daily
2026-04
- Apr 30, 2026 RL Patches 3D Consistency Into Video Models Without Touching Architecture Daily
- Apr 29, 2026 Emotion Probes Crash From 82% to 5% Without Keywords Daily
- Apr 28, 2026 ProEval Cuts Benchmark Eval Samples 8-65x Daily
- Apr 27, 2026 Full Traces Lift Multi-Agent Attribution Accuracy 76% Daily
- Apr 26, 2026 4B Agent on 10K Data, MoE Upcycling Saves 32% Compute Daily
- Apr 25, 2026 Coding Agents Start Cheating by Round 4 Under Score Pressure Daily
- Apr 24, 2026 Recalibrating the Critic Lifts Reasoning Models 18 Points Daily
- Apr 23, 2026 A 305M Retriever Gains 45% on Instruction Following Daily
- Apr 22, 2026 Agents Ignore Answers Placed in Plain Sight Daily
- Apr 21, 2026 3B Matches R1 on Refusal; B Matrix Is LoRA's Bottleneck Daily
- Apr 20, 2026 Open Omni Hits Flagship Scale, Self-Judge Breaks, Reasoning Leaks Forgotten Facts Daily
- Apr 19, 2026 Compile the Corpus Into a Skill Tree, Train Surrogates on Logs Daily
- Apr 18, 2026 Tencent Open-Sources 3D World Generation, VLM Modal Bias Probe Daily
- Apr 17, 2026 Big Models Resist Rumors but Fall for Noise Daily
- Apr 16, 2026 VLMs Break When You Change the Rules Daily
- Apr 15, 2026 dLLMs Hallucinate Differently, PRM Labeling Cost Drops 100x Daily
- Apr 14, 2026 SFT Convergence Hides Failures, Attention Hijacking Hits 94% Daily
- Apr 13, 2026 DMax Triples Parallel Decoding Efficiency for Diffusion LMs Daily
- Apr 12, 2026 Scrambled Media Boosts Reasoning; 6B Model Tops GPT-4o Daily
- Apr 11, 2026 1.7x Faster From Fine-Tuning Alone, Token Collapse Misdiagnosed Daily
- Apr 10, 2026 Entropy Is Lying to You, Implicit Reasoning Tops Out at 7 Steps Daily
- Apr 9, 2026 120B on One GPU, and 40% of Video Benchmarks Are Guessable Daily
- Apr 8, 2026 Streaming Video QA Hits 2 FPS, RLVR Shrugs Off Noisy Labels Daily
- Apr 7, 2026 Learned Sparsity Cuts Diffusion Inference Compute by 54% Daily
- Apr 6, 2026 Open-Source 32B Cracks Hardware Code, Agents Score Just 23% Daily
- Apr 5, 2026 4M Game Frames Train Rendering, Internalized Skills Beat Retrieval Daily
- Apr 4, 2026 Single Neurons Remember Entities, Reusable Routines Boost 19% Daily
- Apr 3, 2026 Minimalist Agents Match MCP, Code Models Think Mid-Stream Daily
- Apr 2, 2026 Data Mixing Becomes Post-Training, Surface Cues Hijack Reasoning 38x Daily
2026-03
- Mar 30, 2026 Watermarks Enable Bit-Level Tracing, Diffusion VLMs Ground GUI Daily
- Mar 29, 2026 Mistral Ships TTS, Diffusion LLMs Get 4.7x Faster Daily
- Mar 28, 2026 Self-Distillation Strips Out Hesitation, OOD Drops 40% Daily
- Mar 27, 2026 Speculative Execution Hits Agent Loops, 3x Faster Daily
- Mar 26, 2026 Diffusion OCR Decodes 3.2x Faster, Single-Stream AV in 2 Seconds Daily
- Mar 25, 2026 PDEs Beat Attention 2x, Local RL Saves 3/4 Compute Daily
- Mar 24, 2026 Seed1.8 Goes Agent-Native, Language Training Erodes Vision Daily
- Mar 23, 2026 12B Beats GPT-4, Distilled Students Surpass Teachers Daily
- Mar 22, 2026 3B Params Win Three Olympiad Golds, 768-D Discrete Tokens Work Daily
- Mar 21, 2026 3D at 0.1% Tokens, Video Fine-Tuning's Hidden Spatial Cost Daily
- Mar 20, 2026 First 32B Industrial Code Model, War-Tested Reasoning Eval Daily
- Mar 19, 2026 Open-Source Search Agent Wins With 12K Samples, Agent Skills Mostly Fail Daily
- Mar 18, 2026 700K Paper Pairs Distill Taste, Null Spaces Expose Blind Spots Daily
- Mar 17, 2026 Expert Reasoning Structure for CoT, +13% on Novel Class Discovery Daily
- Mar 16, 2026 Budget-Aware Agents Beat 4x Brute-Force Sampling Daily
- Mar 15, 2026 Document Agents Navigate by Luck, Prefill Speeds Up 1.82x Daily
- Mar 14, 2026 Encode the Answer, Not the Question — Embeddings Gain 9% Daily
- Mar 13, 2026 \"Think It Over\" Can Unlock a Model's Memory Bank Daily
- Mar 12, 2026 Write Code Before You Draw, Layouts Improve 68% Daily
- Mar 11, 2026 4-Step Diffusion Beats 100-Step Baselines, Layer Skipping Saves 18% Daily
- Mar 10, 2026 12k Samples Beat Finance SOTA, CUDA Optimization 35% Faster Daily
- Mar 9, 2026 Drop CLIP, Gain Performance: VLMs Work Better Without It Daily
- Mar 8, 2026 \"Be Concise\" Halves Tokens, Lifts Accuracy by 16 Points Daily
- Mar 7, 2026 14B Video Model Runs Real-Time on a Single GPU Daily
- Mar 6, 2026 Code Agents Can't Cross Repo Boundaries, Under 45% Success Daily
- Mar 5, 2026 Direct Lottie Generation, DPO's Built-In Forgetting Defense Daily
- Mar 4, 2026 9K Samples Rival R1, Most RL Gains Trace Back to SFT Daily
- Mar 3, 2026 Spectral Conditions Unify μP Scaling, Data Curation Leaks Privacy Daily
- Mar 2, 2026 Drop 90% of Vision Tokens, Keep the Performance Daily
- Mar 1, 2026 Latent Reasoning's Gains Aren't From Reasoning Daily
2026-02
- Feb 28, 2026 Tri-Modal Training From Scratch, Agentic RL Gets a Stability Fix Daily
- Feb 27, 2026 TTT Is Linear Attention, Terminal Agent Data Recipe Goes Open Daily
- Feb 26, 2026 11 Agent Failure Modes From Red-Teaming, Step-Level Routing Cuts Cost 700x Daily
- Feb 25, 2026 Token Probabilities as Zero-Shot Rewards Hit 0.95 Correlation Daily
- Feb 24, 2026 74% of Agent Coordination May Be Wasted Effort Daily
- Feb 23, 2026 Model Folding Beats Pruning, XR Gets Hand-Level Control Daily
- Feb 22, 2026 Adaptive DiT Patches Hit 3x Speedup, Mamba Improves by Subtraction Daily
- Feb 21, 2026 Agents Score Higher but Fail the Same Way Daily
- Feb 20, 2026 Example Pairs Replace Prompts, Agents Play Favorites Daily
- Feb 19, 2026 Spectral Decay Recovers 7% Accuracy in W4A4 Quantization Daily
- Feb 18, 2026 Binary Tokens Make Image Gen 30x Faster, RL Training Learns to Reflect Daily
- Feb 17, 2026 Online RL Cracks Web Agents, Reward Models Learn to Look Backward Daily
- Feb 16, 2026 Vertical AI Is Winning: Medical, Robotics, and Science Agents Daily
- Feb 15, 2026 Running Out of RL Training Data? Just Combine the Easy Problems Daily
- Feb 14, 2026 11B Active Parameters Hit Frontier-Level Agent Intelligence Daily
- Feb 13, 2026 AI Solves Real Open Math Problems, World Models Everywhere Daily
- Feb 12, 2026 Text Diffusion Hits Practical Speed, RL Spreads Everywhere Daily
- Feb 11, 2026 Agent Bottlenecks Are Shifting From Models to Systems Daily
- Feb 10, 2026 LinkedIn Ships LLM-Powered Search Ranking at Scale Daily
- Feb 9, 2026 Medical LLMs Should Ask Questions, Not Just Answer Them Daily
- Feb 8, 2026 Diffusion Drafting Hits 6x Speedup, 14B Beats Claude at Kernels Daily
- Feb 7, 2026 Trillion-Parameter Multimodal, 4B Agents Match 671B, PPO Exposed Daily
- Feb 6, 2026 256 Tokens Match Full Attention, Agents That Build Agents Daily
- Feb 5, 2026 Kimi K2.5 Open-Sources Agent Swarm, CoT Plans Only 2-3 Steps Ahead Daily
- Feb 4, 2026 Better SFT Makes Worse RL, Distillation Waste, Reward Circuits Daily
- Feb 3, 2026 Zero-Cost Data Mix Search, Guided RLVR, Selective SFT Daily
- Feb 2, 2026 Unlimited RLVR Data From Web Text, FP4 Pretraining Matches BF16 Daily
- Feb 1, 2026 Open-Source Deep Research Beats GPT-5, Embedding Scaling Outshines Experts Daily