Morph MORPH
SIGN IN SIGN UP
dhruvhead / jev-exploration UNCLAIMED
0 0 2 Python

COMMITS

main
September 18, 2026
S
Merge pull request #8 from SamuelSacco/claude/jev-benchmarks-review-n555nl
Samuel Sacco committed
af91a6e
S
Issue #5: run negation probe (1 pass)
Samuel Sacco committed
e394b43
C
Issue #1: analyse the gradient run — the ECE rise is mass, not calibration
Claude committed
b6330cc
S
Issue #1: run difficulty-gradient experiment (3 passes, 60 calls)
Samuel Sacco committed
ec38dc8
C
Issue #1: build and freeze the difficulty-gradient dataset
Claude committed
be69a9a
C
Issue #5: retire the rerank demo, replace it with a negation probe
Claude committed
cd7a2d7
C
Issue #6: settle the two disputed triage labels, and version the label set
Claude committed
65fbc0d
C
Issue #2: recompute every published Jev ECE against its own noise floor
Claude committed
ae9fea0
C
Make measure_network_floor() proxy-aware, and split the floor into phases
Claude committed
c02cd51
S
Issues #3, #4: re-run demos x5 on new harness, measure network floor, update ledger
Samuel Sacco committed
dfbeb52
September 17, 2026
C
Link the open issues from the README, with the reporting standard
Claude committed
0aab5b8
C
Add 52 offline tests and CI
Claude committed
103112a
C
Rebuild the lab harness so its numbers can be audited
Claude committed
8112ca5
C
Add calibration analysis, and correct the ledger with an ECE noise floor
Claude committed
03458fd
C
Collapse three HTTP clients into one, with the retries the docs require
Claude committed
aba6529
C
Reframe docs as a maintained evidence ledger
Claude committed
fcdb430
C
Add critical review of repo vs public Jev benchmarks, plus baselines
Claude committed
f991859
S
Make verifier + skill portable: direct API via TYPESAFE_API_KEY, add .gitignore
Samuel Sacco committed
1d28a38
S
Jev exploration: thread, claim audit, demos, verifier prototype
Samuel Sacco committed
b58ddfa