Sakana AI
Researchers introduced Transformer², a machine learning system that dynamically adjusts its weights for different tasks using Singular Value Decomposition and reinforcement learning. The method learns task-specific z-vectors that modulate weight matrix components, requiring far fewer parameters than LoRA while achieving comparable or better performance on math, coding, reasoning, and visual tasks. This approach enables LLMs to adapt to new tasks at inference time without retraining, and z-vectors learned on one model can partially transfer to another model.
Sakana AI
● 5 sources
Sakana AI released a reasoning benchmark based on Sudoku puzzles to test and improve AI models' logical reasoning capabilities. The benchmark includes thousands of curated traditional and modern Sudoku puzzles, with data extracted from thousands of hours of reasoning explanations from YouTube channel Cracking The Cryptic, where world-championship-level solvers narrate their step-by-step solving process. Current state-of-the-art models struggle significantly, with only OpenAI's o3 achieving a 5% success rate on the easiest puzzles, highlighting the gap between human-like reasoning and contemporary AI approaches.
TLDR Dev
·
19 hours ago
● 5 sources
Kimi K3, a Chinese AI model, delivers comparable code quality to Claude at three-to-five times lower API costs and more generous subscription tiers, while also operating without US restrictions that hobble Anthropic's offerings. K3's API pricing is $3 per million input tokens and $15 per million output, compared to Claude's $10 and $50, with Kimi's $39 monthly tier significantly outperforming Claude's metered plans. The price and capability gap suggests US regulatory policy has backfired by constraining American customers while leaving unrestricted Chinese alternatives accessible, potentially reshaping which AI models dominate the market.
Apple ML Research
·
1 day ago
Researchers introduced RayRoPE, a positional encoding method for multi-view transformers that represents patch positions using predicted 3D points along camera rays rather than ray directions. The method achieves 15% relative improvement on LPIPS metrics in the CO3D dataset for novel-view synthesis compared to alternative position encoding schemes. RayRoPE enables geometry-aware attention that maintains SE(3) invariance and can incorporate RGB-D inputs, improving performance on multi-view 3D reconstruction tasks.
