Agents Face Long Horizons and Better Training Data

Featured

• FACET: Preserving Source Intent and Executable State in Terminal Task Synthesis score 10
  入选 HF Daily Papers; HF 热度: 109 upvotes (+4); 有代码实现; 关键词(2): fine-tuning, agentic
• SkillForge: Self-Distilling Agents for Project-Specific Issue Resolution score 7
  入选 HF Daily Papers; HF 热度: 11 upvotes (+3); 关键词(2): distillation, open-source
• FM-Bench: A Benchmark for Long-Horizon Management with Competing Agents score 8
  入选 HF Daily Papers; HF 热度: 16 upvotes (+3); 有代码实现

Also Worth Noting

• Training Chemical Plausibility-Aware Large Language Models for Single-Step Retrosynthesis score 8
  入选 HF Daily Papers; HF 热度: 30 upvotes (+4); 关键词(1): fine-tuning
• SPADE: Self-Play in Adaptive Synthetic Executable Environments score 10
  入选 HF Daily Papers; HF 热度: 45 upvotes (+4); 有代码实现; 关键词(6): scaling, edge, pretraining, agentic, tool use
• TranslatePsy-AfriSLM: High-Quality Data Scaling For Low-Resource Machine Translation score 4
  机构: Meta; 关键词(3): scaling, open-source, synthetic data