-
v0.1.12: full replicate directories for audit_cb_* and temp_il_* (v0.1.11 upload was incomplete); known-truth simulation
-
v0.1.11: counterbalanced end-to-end check, interleaved temperature check
-
v0.1.10: end-to-end check runs, interleaved hosted dispatch runs, v4 analysis outputs
-
v0.1.8: isolated-call sensitivity arm for endpoint A, crossed-bootstrap outputs, new paper title
-
Update card for release 0.1.7; add v3 analysis outputs and provenance for the logged local-model reruns
-
Add floor16v3_phi3_t0: logged rerun with attempt-level logs
-
Add floor16v3_phi3: logged rerun with attempt-level logs
-
Add floor16v3_qwen3_t0: logged rerun with attempt-level logs
-
Add floor16v3_qwen3: logged rerun with attempt-level logs
-
Add floor16v3_mistral_t0: logged rerun with attempt-level logs
-
Add floor16v3_mistral: logged rerun with attempt-level logs
-
Add floor16v3_llama_t0: logged rerun with attempt-level logs
-
Add floor16v3_llama: logged rerun with attempt-level logs
-
Paired hosted comparison
-
Add hosted temperature-1.0 arm
-
Update temperature comparison with hosted check
-
Add hosted temperature-0 check (Groq, gpt-oss-20b)
-
Add temperature_compare.json
-
Add robustness.json
-
Add floor16_qwen3_t0 (temperature 0, five replicates)
-
Add floor16_mistral_t0 (temperature 0, five replicates)
-
Add floor16_ollama_t0 (temperature 0, five replicates)
-
Add floor16_phi3_t0 (temperature 0, five replicates)
-
Add floor16_phi3 trajectories (ten replicates, sixteen vignettes)
-
Add floor16_qwen3 trajectories (ten replicates, sixteen vignettes)
-
Add floor16_mistral trajectories (ten replicates, sixteen vignettes)
-
Add floor16_ollama trajectories (ten replicates, sixteen vignettes)
-
FairMedAgent v0.1.0: instability-floor trajectories, datasheet, adjudication record