PyTorch toolkit for probing which intermediate feature depth of a frozen backbone works best for a downstream task.
Safe low-rate NVIDIA Build free model availability probe. Tests which NVIDIA API models work from your server/IP region, measures latency, enriches model metadata, and exports JSON/CSV/Excel reports.
Predicting LLM answer correctness using hidden-state and uncertainty probes across 424K examples and five QA/reasoning benchmarks.
Probe once. Configure everywhere. Evidence-first model capability probing and configuration for coding agents.
A probing framework for Speech Emotion Recognition (SER) transformers to evaluate how and where paralinguistic, acoustic knowledge is encoded across their hidden layers.
To associate your repository with the model-probing topic, visit your repo's landing page and select "manage topics."