1
I built an LLM debugger for fine-tuning failures
为LoRA微调失败找出元凶，用训练数据归因精准定位回归样本。
Article URL: https://github.com/gradian-ai/gradian Comments URL: https://news.ycombinator.com/item?id=49168696 Points: 3 # Comments: 3
2
DRIFT: Refining Instruction Data via On-Policy Data Attribution
新方法DRIFT通过on-policy数据归因，精准优化SFT训练数据分布，助力突破大模型能力上限。
arXiv:2606.18307v1 Announce Type: new Abstract: Optimizing the training data distribution for Supervised Fine-Tuning (SFT) dictates the capability of …
3
Mechanistic Data Attribution: Tracing the Training Origins of Interpretable LLM Units
追溯大语言模型中可解释单元的训练数据源头，为理解模型行为提供全新归因方法
arXiv:2601.21996v2 Announce Type: replace-cross Abstract: While Mechanistic Interpretability has identified interpretable circuits in LLMs, their caus…
4
STRIDE: Training Data Attribution via Sparse Recovery from Subset Perturbations
用稀疏恢复技术从子集扰动中精确归因训练数据，破解大模型因果干预的计算难题
arXiv:2606.05165v1 Announce Type: cross Abstract: Training Data Attribution (TDA) seeks to trace a model's predictions back to its training data. The …
5
On the Fragility of Data Attribution When Learning Is Distributed
分布式学习中数据归因的脆弱性：单个参与者可操纵归因值大幅膨胀，挑战定价与审计可信度。
arXiv:2605.15520v1 Announce Type: cross Abstract: Data attribution has become an important component of pricing, auditing, and governance in machine l…