method · first seen 4 Oct 2026
model soup
Also: weight averaging, weight snapshot averaging
Merging weights of several fine-tuned models with the same initialization to improve performance without extra inference cost.
Mentions per week
PapersCodeDiscussion
Earliest sources
- Understanding the Weight Averaging Mechanism in LLM Training for Post-Training QuantizationPapers · 4 Oct 2026 · Hanzhang Wang, Tianqi Shen, Zonglin Liu, Junze He, Difan Zou, Ziye Ma
- Data Curation and Model Souping: Unlocking Qwen3-0.6B for MS MARCO Passage RankingPapers · 10 Oct 2026 · Shuguang Han
- Sparse Recovery Bird's-Eye View Model Soup for Robust Camera-Based 3D Object DetectionPapers · 26 Oct 2026 · Jinwoo Chung, Sangho An, Hyunsung Park, Jangho Kim
- GradReason: LLMs as Meta-Learners that Replace Training Epochs with Training-Data-Free Weight UpdatesPapers · 7 Nov 2026 · Ali Mohammadi Ruzbahani, Abbas Yazdinejad, Hadis Karimipour
Latest sources
- GradReason: LLMs as Meta-Learners that Replace Training Epochs with Training-Data-Free Weight UpdatesPapers · 7 Nov 2026 · Ali Mohammadi Ruzbahani, Abbas Yazdinejad, Hadis Karimipour
- Sparse Recovery Bird's-Eye View Model Soup for Robust Camera-Based 3D Object DetectionPapers · 26 Oct 2026 · Jinwoo Chung, Sangho An, Hyunsung Park, Jangho Kim
- Data Curation and Model Souping: Unlocking Qwen3-0.6B for MS MARCO Passage RankingPapers · 10 Oct 2026 · Shuguang Han
- Understanding the Weight Averaging Mechanism in LLM Training for Post-Training QuantizationPapers · 4 Oct 2026 · Hanzhang Wang, Tianqi Shen, Zonglin Liu, Junze He, Difan Zou, Ziye Ma