GRACE: Gradient-aligned Reasoning Data Curation for Efficient Post-training
Existing reasoning data curation pipelines score whole samples, treating every intermediate step as equally valuable. In reality, steps within a trace contribute very unevenly, and selecting reasoning data well requires assessing them individually. We present GRACE, a gradient-aligned curation method that views each reasoning trace as a sequence of optimization events and scores every step by two complementary signals: its alignment with the answer-oriented gradient direction, and its consistenc
Record details
Published: 13 May 2026
Source: arXiv
Category: Research
Topics: Safety & alignment
Retrieved: 14 July 2026
Related evidence
These records share source-supplied organisations, an exact publisher byline, automatic topics or regions. The reason is shown on every link; related does not mean supporting, agreeing with or verifying this record.
Exact Linear Attention
arXiv · 13 May 2026
STAR: Semantic-Temporal Adaptive Representation Learning for Few-Shot Action Recognition
arXiv · 13 May 2026
Improving Code Translation with Syntax-Guided and Semantic-aware Preference Optimization
arXiv · 13 May 2026
Revealing the Gap in Human and VLM Scene Perception through Counterfactual Semantic Saliency
arXiv · 13 May 2026
Tracing Persona Vectors Through LLM Pretraining
arXiv · 13 May 2026
Not Just RLHF: Why Alignment Alone Won't Fix Multi-Agent Sycophancy
arXiv · 13 May 2026
How to cite this record
ethics.ai (13 May 2026), “GRACE: Gradient-aligned Reasoning Data Curation for Efficient Post-training,” evidence record 4407, https://ethics.ai/record/4407 (originally published by arXiv).
Use and limitations
This page is a stable index and citation surface for a source record. ethics.ai did not author the underlying report and has not independently verified every claim. Automatic topics may be imperfect. For consequential use, quote and cite the original publisher.