view post Post 1,988 Uhh did Opus 4.8 cheat on PostTrainBench??it found an API key in the PostTrainBench environment that allowed it to generate synthetic training data without using GPU hours, boosting the base model by 0.4913Source: https://posttrainbench.com/traces/run.html?id=claude_non_api_max_claude-opus-4-8_10h_run1__healthbench_Qwen_Qwen3-4B-Base_17315102#tab=trace See translation 1 reply · 👀 3 3 🔥 1 1 + Reply
DynaFLIP: Rethinking Robotics Perception via Tri-Modal-Dynamics Guided Representation Paper • 2605.30350 • Published May 28 • 10
Contrastive Distribution Matching for Amortized Sequential Monte Carlo in Discrete Diffusion Paper • 2605.23346 • Published May 22
optimize_anything: A Universal API for Optimizing any Text Parameter Paper • 2605.19633 • Published May 19 • 6
MultiGen: Level-Design for Editable Multiplayer Worlds in Diffusion Game Engines Paper • 2603.06679 • Published Mar 30 • 5
AVO: Agentic Variation Operators for Autonomous Evolutionary Search Paper • 2603.24517 • Published Mar 25 • 14
V-Co: A Closer Look at Visual Representation Alignment via Co-Denoising Paper • 2603.16792 • Published Mar 17 • 3
SE-Bench: Benchmarking Self-Evolution with Knowledge Internalization Paper • 2602.04811 • Published Feb 4 • 2
UM-Text: A Unified Multimodal Model for Image Understanding Paper • 2601.08321 • Published Jan 13 • 21
ResTok: Learning Hierarchical Residuals in 1D Visual Tokenizers for Autoregressive Image Generation Paper • 2601.03955 • Published Jan 7 • 3