{"slug": "forge-forward-only-test-time-adaptation-for-integer-only-vision-models-on", "title": "FORGE: Forward-Only Test-Time Adaptation for Integer-Only Vision Models on Microcontrollers", "summary": "Researchers introduced FORGE, a forward-only test-time adaptation method for integer-only vision models on microcontrollers, recovering 93% of gradient-based TENT's accuracy gain (+20.9 vs. +24.9 points) while being the only method that runs on folded integer-only models. Deployed on an ESP32-S3, adaptation costs only 8.3 mJ (6.8% of inference energy) and 21.9 ms, demonstrating feasibility on real hardware.", "body_md": "arXiv:2609.01683v1 Announce Type: new\nAbstract: Vision models deployed on microcontrollers (MCUs) are quantized to integer-only arithmetic and run in inference-only runtimes that do not carry the machinery backpropagation needs: the standard tool for adapting a model to the distribution shift (sensor noise, blur, lighting) it meets in the field. Existing forward-only test-time adaptation (TTA) methods either run only on server- or edge-GPU-class models (not true microcontroller integer execution), or require the batch-normalization (BN) layers that integer deployment fuses away. We present a forward-only TTA method that operates on deployed, BN-folded, integer-only convolutional networks. The key observation is that fusing BN into the preceding convolution, a mandatory step for integer inference, destroys the statistics that normalization-based adaptation relies on. We restore adaptation by re-normalizing each folded convolution's per-channel output to its clean training statistics, using only forward-pass estimates. The method (i) recovers most of gradient-based TENT's accuracy gain (+20.9 vs. +24.9 points) and matches forward-only BN adaptation, while being the only method that runs on a folded integer-only model; (ii) needs to adapt only 3 of 21 layers (selected without seeing the test corruptions) to recover 93% of the benefit; (iii) survives single-sample streaming with a batch-size-scaled momentum; and (iv) generalizes across three datasets (up to 200 classes) and two architectures. We validate bit-exact int8 convolution execution and deploy on an ESP32-S3, where, measured with a Nordic PPK2 power profiler, the forward-only adaptation (a lightweight fp32 recalibration around the int8 convolutions) costs only 8.3 mJ (6.8% of inference energy) and 21.9 ms on the deployed SIMD-optimized model: forward-only adaptation is cheap on a real microcontroller.", "url": "https://wpnews.pro/news/forge-forward-only-test-time-adaptation-for-integer-only-vision-models-on", "canonical_source": "https://arxiv.org/abs/2609.01683", "published_at": "2026-09-03 04:00:00+00:00", "updated_at": "2026-09-03 04:22:36.435408+00:00", "lang": "en", "topics": ["artificial-intelligence", "machine-learning", "computer-vision"], "entities": ["FORGE", "TENT", "ESP32-S3", "Nordic PPK2"], "alternates": {"html": "https://wpnews.pro/news/forge-forward-only-test-time-adaptation-for-integer-only-vision-models-on", "markdown": "https://wpnews.pro/news/forge-forward-only-test-time-adaptation-for-integer-only-vision-models-on.md", "text": "https://wpnews.pro/news/forge-forward-only-test-time-adaptation-for-integer-only-vision-models-on.txt", "jsonld": "https://wpnews.pro/news/forge-forward-only-test-time-adaptation-for-integer-only-vision-models-on.jsonld"}}