Zero-Cost Data Mix Search, Guided RLVR, Selective SFT

Featured

• Decouple Searching from Training: Scaling Data Mixing via Model Merging for Large Language Model Pre-training score 7
  入选 HF Daily Papers; HF 热度: 8 upvotes (+2); 关键词(7): scaling, efficiency, pre-training, search, benchmark
• Adaptive Ability Decomposing for Unlocking Large Reasoning Model Effective Reinforcement Learning score 7
  入选 HF Daily Papers; HF 热度: 5 upvotes (+2); 关键词(2): distillation, reasoning
• Pareto-Conditioned Diffusion Models for Offline Multi-Objective Optimization score 7
  机构: Google; 关键词(1): diffusion; 顶会接收: ICLR
• RecGOAT: Graph Optimal Adaptive Transport for LLM-Enhanced Multimodal Recommendation with Dual Semantic Alignment score 5
  入选 HF Daily Papers; 关键词(6): deployment, alignment, attention, reasoning, multimodal
• Segment-Level Attribution for Selective Learning of Long Reasoning Traces score 5
  关键词(3): efficiency, finetuning, reasoning; 顶会接收: ICLR
• DuoGen: Towards General Purpose Interleaved Multimodal Generation score 5
  机构: NVIDIA; 关键词(11): alignment, pretraining, transformer, diffusion, reasoning
• How Far Are LLMs from Professional Poker Players? Revisiting Game-Theoretic Reasoning with Agentic Tool Use score 5
  关键词(3): agentic, tool use, reasoning; 顶会接收: ICLR
• Kernelized Edge Attention: Addressing Semantic Attention Blurring in Temporal Graph Neural Networks score 5
  关键词(3): edge, transformer, attention; 顶会接收: AAAI

Also Worth Noting

• Bridging Degradation Discrimination and Generation for Universal Image Restoration score 4
  关键词(1): diffusion; 顶会接收: ICLR
• Topology and Geometry of the Learning Space of ReLU Networks: Connectivity and Singularities score 4
  关键词(1): pruning; 顶会接收: ICLR
• Alignment of Diffusion Model and Flow Matching for Text-to-Image Generation score 2
  关键词(6): fine-tuning, finetuning, alignment, diffusion, text-to-image
• Toward Autonomous Laboratory Safety Monitoring with Vision Language Models: Learning to See Hazards Through Scene Structure score 2
  关键词(5): alignment, post-training, reasoning, evaluation, safety
• PolarMem: A Training-Free Polarized Latent Graph Memory for Verifiable Multimodal Agents score 2
  关键词(7): inference, agent, agents, reasoning, multimodal
• Text is All You Need for Vision-Language Model Jailbreaking score 2
  关键词(5): alignment, multimodal, vision-language, safety, jailbreak
• Open Materials Generation with Inference-Time Reinforcement Learning score 2
  关键词(2): efficiency, inference
• LLMs as High-Dimensional Nonlinear Autoregressive Models with Attention: Training, Alignment and Inference score 2
  关键词(11): serving, inference, fine-tuning, RLHF, DPO
• When Agents "Misremember" Collectively: Exploring the Mandela Effect in LLM-based Multi-Agent Systems score 2
  关键词(4): alignment, agent, agents, benchmark
• Exact Instance Compression for Convex Empirical Risk Minimization via Color Refinement score 2
  关键词(2): scaling, compression
• DISK: Dynamic Inference SKipping for World Models score 2
  关键词(4): inference, diffusion, planning, cost
• IceBench-S2S: A Benchmark of Deep Learning for Challenging Subseasonal-to-Seasonal Daily Arctic Sea Ice Forecasting in Deep Latent Space score 2
  关键词(5): efficiency, deployment, planning, benchmark, evaluation
• Towards Building Non-Fine-Tunable Foundation Models score 2
  关键词(4): fine-tuning, pre-training, pretraining, safety
• Do Latent-CoT Models Think Step-by-Step? A Mechanistic Study on Sequential Reasoning Tasks score 2
  关键词(3): distillation, attention, reasoning
• Model Optimization for Multi-Camera 3D Detection and Tracking score 2
  关键词(8): quantization, latency, fine-tuning, post-training, pretraining