SACQ: Structure-Aware Codebook Quantization for Ultra-Low-Bit Vision-Language-Action Models
SACQ: Structure-Aware Codebook Quantization for Ultra-Low-Bit Vision-Language-Action Models
This paper introduces SACQ, a weight-only vector-codebook quantization framework designed for Vision-Language-Action (VLA) models to enable deployment on resource-constrained robotic platforms. The method employs structure-aware codebook fitting, action-aware mixed-bit allocation, and two-stage post-quantization refinement. Evaluated on the GR00T N1.7 model, SACQ achieves 93.85% average success across four LIBERO suites at 1.5 bits per language/action weight, reducing memory footprint by 4.83x. It also retains 93.55% success with INT8 activation quantization while running 1.51x faster than BF16. On the SimplER benchmark, it achieves 63.76% and 52.27% success under Visual Matching and Variant Aggregation at 2 bits per weight, outperforming BF16 baselines.
The authors present SACQ, a quantization framework for Vision-Language-Action models that claims to maintain high success rates on the LIBERO and SimplER benchmarks while significantly reducing memory usage and improving inference speed compared to BF16 baselines.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
memory reduction factor
4.83
View original evidence
el memory by 4.83x. With INT8 activation quantization and low-precision attention, it retains 93.55% success while running 1.51x faster than BF16. On SimplER, SACQ further achieves 63.76% and 52.27% success under Visual Matching and Variant Aggregation at 2 bits per language/action weight, compared to 58.90% and 53.45% for BF16, respectively.
Open source S5speedup factor
1.51
View original evidence
el memory by 4.83x. With INT8 activation quantization and low-precision attention, it retains 93.55% success while running 1.51x faster than BF16. On SimplER, SACQ further achieves 63.76% and 52.27% success under Visual Matching and Variant Aggregation at 2 bits per language/action weight, compared to 58.90% and 53.45% for BF16, respectively.
Open source S5bits per language/action weight
1.5
View original evidence
allocation based on the increase in final-action error as individual layers are compressed; and (3) two-stage post-quantization refinement that reassigns codeword indices with second-order error compensation and calibrates codebooks and row-group scales against the decoded action. Using GR00T N1.7, SACQ achieves 93.85% average success across four LIBERO suites at 1.5 bits per language/action weight while reducing mod
Open source S4bits per language/action weight
2
View original evidence
el memory by 4.83x. With INT8 activation quantization and low-precision attention, it retains 93.55% success while running 1.51x faster than BF16. On SimplER, SACQ further achieves 63.76% and 52.27% success under Visual Matching and Variant Aggregation at 2 bits per language/action weight, compared to 58.90% and 53.45% for BF16, respectively.
Open source S5
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
allocation based on the increase in final-action error as individual layers are compressed; and (3) two-stage post-quantization refinement that reassigns codeword indices with second-order error compensation and calibrates codebooks and row-group scales against the decoded action. Using GR00T N1.7, SACQ achieves 93.85% average success across four LIBERO suites at 1.5 bits per language/action weight while reducing mod
Open source S4
el memory by 4.83x. With INT8 activation quantization and low-precision attention, it retains 93.55% success while running 1.51x faster than BF16. On SimplER, SACQ further achieves 63.76% and 52.27% success under Visual Matching and Variant Aggregation at 2 bits per language/action weight, compared to 58.90% and 53.45% for BF16, respectively.
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org