From Fewer Samples to Fewer Bits: Reframing Dataset Distillation as Joint Optimization of Precision and Compactness

Dimensions overall score 8.0

GitHub Code Pulse

No public code linked for this paper yet.

• Quantization-aware Dataset Distillation (QuADD) is a unified framework that jointly optimizes dataset compactness and precision under fixed bit budgets.method 95%
• QuADD integrates a differentiable quantization module within the distillation loop for end-to-end co-optimization.method 95%
• The framework supports both uniform and adaptive non-uniform quantization.method 95%
• QuADD surpasses existing Dataset Distillation (DD) and post-quantized baselines in accuracy per bit.result 95%
• The paper empirically analyzes how bit allocation between sample count and precision influences learning performance.method 90%
• Adaptive non-uniform quantization learns quantization levels from data to represent information-dense regions better.technical 90%
• QuADD establishes a new standard for information-efficient dataset distillation.result 85%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card