Provenance
Records where datasets, evaluation results and corrections came from, so each reported finding can be traced back to its inputs.
Research Stack v0.1.0 is public research software connecting TRACE provenance, LOOP evaluation, LENS visualization and FRAME review workspaces. Complete source, NRS-Bench, three technical reports and reproduction instructions are available; software validation remains developer/CI-reported and synthetic.
Conceptual systems illustration, not application data. Research Stack v0.1.0 is public research software; validation is developer/CI-reported. Epistemic Trust Lab read-only import was verified. Resident Model Lab and VEYDRIS integrations are proposed, not completed.
The v0.1.0 Initial Public Research Release includes complete public source, NRS-Bench, versioned synthetic datasets, three technical reports and reproduction instructions. TRACE · LOOP · LENS · FRAME connects evidence provenance, deterministic evaluation, scientific visualization and reviewed corrections. These are controlled synthetic software-validation results, not a real-model benchmark or human study.
Records where datasets, evaluation results and corrections came from, so each reported finding can be traced back to its inputs.
Runs evaluations over a dataset, supports reviewed corrections, and reruns the evaluation after review.
Presents evaluation results and their changes visually for inspection by a human reviewer.
Brings the workflow together and produces a reproducible report of the reviewed results.
35 backend · 12 visualization · 7 browser · developer/CI-reported
Windows, Linux and macOS reproduction · developer/CI-reported
Epistemic Trust Lab data imported read-only; existing projects unchanged · developer reported
Historical context: the November 2, 2026 local-build report recorded 31 passing tests. That baseline is historical; the public v0.1.0 release reports 54 automated tests.
Validation concerns controlled synthetic evaluations and software reproduction, not real-model performance or a human-user study.
No independent-party replication or peer review is claimed.
The static results demonstration is read-only exported output, not the full evaluation application. Install the application to execute evaluations and reviewed corrections.
Resident Model Lab and VEYDRIS are separate programs; integration is proposed, not completed.
Local OLMoE inference with router-faithful expert residency. Compare the v0.2 memory, latency and fidelity findings with their experimental limits.
Recorded-video detection, temporal tracking, replay and exports in a local prototype. The synthetic map is not verified geolocation.
Evidence provenance, conversational memory and inspectable uncertainty, with founder-led Epistemic Trust Lab synthetic-policy evaluation kept distinct.
Public notes disclose reported methods and limitations; they are not a substitute for independent validation.
All research themes →