{"slug": "recti-q-feature-space-rectification-for-out-of-distribution-robust-quantized-in", "title": "Recti-Q: Feature-Space Rectification for Out-of-Distribution-Robust Quantized Perception in Edge Robotics", "summary": "A new study from arXiv (2607.18540v1) reveals that post-training quantization (PTQ) on edge robotics platforms creates a Quantization-Induced Robustness Gap, degrading reliability under distribution shifts like sensor noise and severe weather despite preserving clean in-distribution accuracy. The researchers propose Recti-Q, a lightweight feature-space rectification framework that freezes the quantized backbone and trains a small classifier-head LoRA adapter using only source data, recovering lost robustness with less than 1% parameter overhead (as small as 6 KB) and preserving over 99% of PTQ memory savings.", "body_md": "arXiv:2607.18540v1 Announce Type: new\nAbstract: Robotic perception pipelines increasingly rely on large vision backbones deployed on SWaP-constrained edge platforms, making post-training quantization (PTQ) attractive for real-time inference. However, while PTQ often preserves clean in-distribution accuracy, we show that it can substantially degrade reliability under deployment-relevant distribution shifts (e.g., sensor noise, severe weather, and novel operating environments), creating a Quantization-Induced Robustness Gap. Across foundational vision benchmarks (ImageNet-C and PACS), 4-bit PTQ models exhibit pronounced robustness degradation despite negligible ID accuracy loss. To address this, we propose Recti-Q, a lightweight feature-space rectification framework that freezes the quantized backbone and trains a small classifier-head LoRA adapter using only source data. Recti-Q is architecture-agnostic across CNNs and Transformers, supports efficient teacher-free training, and recovers a significant portion of the lost robustness, in some cases matching or exceeding FP32 performance. At less than 1% parameter overhead (as small as 6 KB), Recti-Q preserves over 99% of PTQ memory savings, adds negligible compute, and enables low-bandwidth Over-The-Air (OTA) resilience patching for deployed robotic fleets operating in unpredictable physical environments.", "url": "https://wpnews.pro/news/recti-q-feature-space-rectification-for-out-of-distribution-robust-quantized-in", "canonical_source": "https://arxiv.org/abs/2607.18540", "published_at": "2026-07-22 04:00:00+00:00", "updated_at": "2026-07-22 04:13:53.847404+00:00", "lang": "en", "topics": ["artificial-intelligence", "computer-vision", "robotics", "ai-research"], "entities": ["arXiv", "Recti-Q", "ImageNet-C", "PACS"], "alternates": {"html": "https://wpnews.pro/news/recti-q-feature-space-rectification-for-out-of-distribution-robust-quantized-in", "markdown": "https://wpnews.pro/news/recti-q-feature-space-rectification-for-out-of-distribution-robust-quantized-in.md", "text": "https://wpnews.pro/news/recti-q-feature-space-rectification-for-out-of-distribution-robust-quantized-in.txt", "jsonld": "https://wpnews.pro/news/recti-q-feature-space-rectification-for-out-of-distribution-robust-quantized-in.jsonld"}}