Repository associated with paper titled "CLP: Finetuning Vision-Language-Action Models Requires Fewer Layers Than You Think"
Contract and claim-boundary tooling for empirical AI-system experiments. Turns retained evidence into falsifiable, condition-specific decisions.
Public empirical record of the Triadic Set, a runtime governance architecture for the transformer forward pass. In-band constraint and restoration for LLM inference, with oracle-verified evaluation programs, engineered fixtures, exit-verdict instrumentation, and checksummed telemetry from 1B through 8B scale.
Why more compute doesn't mean faster AI : The Roofline Model explained, with an interactive calculator and Python workload analyzer
Benchmarking GenAI compute efficiency through context compression and model selection.
An open protocol for purposeful, respectful, and resource-conscious AI use.
Adaptive inference algorithm for transformers inspired by quantum collapse (SR framework)
Tracking State-of-the-Art AI Models and Performance is an open-source dataset documenting AI advancements from the 1950s to today. It includes model details, organizations, compute requirements, and benchmarks. Researchers and developers can analyze trends, compare models, and contribute updates. The dataset is open for collaboration $ fostering AI
CROP: compute-efficient SLM pretraining - loss-gated data staging, selective token replay, 2K->4K context ramp, WSD with compute recycling
White paper and specification for the Decentralized Universal Compute Protocol (DUCP).
Project on systems optimisation - neural network where 'experts' bid on compute based on predicted loss.
KAYL (كَيْل): an agent that answers Arabic AI tasks with the least compute. It picks the model and reasoning effort, checks the answer format, and escalates only on failure.
To associate your repository with the compute-efficiency topic, visit your repo's landing page and select "manage topics."