Repository navigation

Complete WebArena-Verified Chromium and Moli trajectory replay - #32

Merged

lanyue-llk merged 122 commits into

mainfrom

codex/webarena-verified-full

Sep 29, 2026

Merged

Complete WebArena-Verified Chromium and Moli trajectory replay#32
lanyue-llk merged 122 commits into
mainfrom
codex/webarena-verified-full

Commits on Sep 29, 2026
- Prepare full WebArena Chromium replay split and evidence ledger
  lanyue-llk
  committed
  4aceaca
- Schedule authored trajectories for uncovered WebArena cases
  lanyue-llk
  committed
  ef89718
- Keep full replay sweep moving after candidate setup errors
  lanyue-llk
  committed
  5875359
- Fix direct replay CLI import and audit first-sweep evidence
  lanyue-llk
  committed
  301b5d1
- Cap structured-target waits by replay action budget
  lanyue-llk
  committed
  66aefc4
- Define uncovered WebArena Chromium trajectories
  lanyue-llk
  committed
  2a3f8bf
- Select walking OSRM engine in authored map cases
  lanyue-llk
  committed
  46a9a93
- Queue WebArena repairs without replaying passes
  lanyue-llk
  committed
  a8fe56d
- Bind Nolan replay to frozen Wikipedia source
  lanyue-llk
  committed
  32f77ae
- Group unresolved WebArena cases by observed failure
  lanyue-llk
  committed
  7bd6225
- Bind WebArena failure groups to source evidence
  lanyue-llk
  committed
  751b837
- Require audited pass counts before repair scheduling
  lanyue-llk
  committed
  8e9a969
- Build WebArena navigation repair paths
  lanyue-llk
  committed
  9d5cc64
- Wait for replay network responses before audit
  lanyue-llk
  committed
  09aa82d
- Preserve WebArena navigation query parameters
  lanyue-llk
  committed
  99c5e52
- Build deterministic OSRM route replay paths
  lanyue-llk
  committed
  68f5df3
- Correct WebArena OSRM route evaluation
  lanyue-llk
  committed
  489a1f1
- Fix WebArena navigation XHR selection
  lanyue-llk
  committed
  aa7b25e
- Fix repeated GitLab member event evaluation
  lanyue-llk
  committed
  6b3a8d4
- Support evidenced action-not-allowed responses
  lanyue-llk
  committed
  c4dafcf
- Fix WebArena markdown URL evaluation
  lanyue-llk
  committed
  2cfe9f5
- Derender URL-valued WebArena form fields
  lanyue-llk
  committed
  516ed67
- Correct unsubmitted coupon form evaluation
  lanyue-llk
  committed
  1b39679
- Handle proxied WebArena transport changes
  lanyue-llk
  committed
  0aa24e8
- Match repeated WebArena network mutations
  lanyue-llk
  committed
  1b5c464
- Finish first Chromium WebArena replay repairs
  lanyue-llk
  committed
  f0f3a72
- Add full WebArena replay catalog runner
  lanyue-llk
  committed
  cc9c135
- Wait for delayed replay navigation
  lanyue-llk
  committed
  5d073ec
- Support recorded waits in portable WebArena trajectories
  lanyue-llk
  committed
  0100f1a
- Audit ordered full WebArena acceptance and targeted repairs
  lanyue-llk
  committed
  6104727
- Bind frozen WebArena replay catalog to replacement site slice
  lanyue-llk
  committed
  bfc2a23
- Bind targeted replay repairs across isolated site preparations
  lanyue-llk
  committed
  2cdcfcb
- Account for duplicate diagnostic execution without replacing first pass
  lanyue-llk
  committed
  7a20812
- Repair overlapping GitLab star replay trajectories
  lanyue-llk
  committed
  811fdc6
- Repair navigation-sensitive replay and cross-deployment bindings
  lanyue-llk
  committed
  eea0e3b
- Repair remaining WebArena Chromium replay trajectories
  lanyue-llk
  committed
  4a1323a
- Build audited full WebArena Chromium replay report
  lanyue-llk
  committed
  4d20401
- Support CowAgent-free Playwright replay backend
  lanyue-llk
  committed
  21180fc
- Disclose unverified data reset in full replay report
  lanyue-llk
  committed
  9a14e2c
- Bind accepted replay trajectories across source deployments
  lanyue-llk
  committed
  eed8feb
- Rebind qualified replay URLs from accepted deployments
  lanyue-llk
  committed
  c09f905
- Bind embedded retrieval replay to its recorded deployment
  lanyue-llk
  committed
  cc0d292
- Bind clean 812-case Chromium acceptance report
  lanyue-llk
  committed
  4a5cbb2
- Identify clean Chromium run as second-round acceptance
  lanyue-llk
  committed
  4204cab
- Bind published Chromium acceptance report
  lanyue-llk
  committed
  83368fb
- fix: observed_permission_denied rejects all-None access_levels
  
  authored and
  lanyue-llk
  committed
  6e04763
- Refactor full WebArena replay pipeline
  lanyue-llk
  committed
  5963dec
- Harden replay repository contracts
  lanyue-llk
  committed
  601851b
- Bind replay releases to maintained WebArena fork
  lanyue-llk
  committed
  2389664
- Harden catalog replay preparation bindings
  lanyue-llk
  committed
  0525754
- Generalize the WebArena catalog pipeline
  lanyue-llk
  committed
  6a0fa47
- Make catalog size a data contract
  lanyue-llk
  committed
  2223875
- Replay only unresolved catalog cases
  lanyue-llk
  committed
  4acf017
- Bind Chromium to the Rainbond ingress
  lanyue-llk
  committed
  34e30b5
- Index catalog rows by task identity
  lanyue-llk
  committed
  fe2fed6
- Resolve catalog runner revision from its checkout
  lanyue-llk
  committed
  ee2375e
- Bind unresolved replay selections as data
  lanyue-llk
  committed
  09f3f3e
- Generalize catalog contracts and pin evaluator provenance
  lanyue-llk
  committed
  8f5c590
- Align Playwright replay execution semantics
  lanyue-llk
  committed
  b7e27fe
- Bind replay action amendments to catalogs
  lanyue-llk
  committed
  90846fe
- Audit replay script execution modes
  lanyue-llk
  committed
  e89135a
- Preserve complete replay retrieval evidence
  lanyue-llk
  committed
  a5a81a3
- Publish Playwright Chromium acceptance report
  lanyue-llk
  committed
  18591de
- Bind replay evaluator to WebArena fork PR
  lanyue-llk
  committed
  72c608c
- Bind task 760 live route contract
  lanyue-llk
  committed
  e21512b
- Publish Moli layout before Playwright actions
  lanyue-llk
  committed
  552337d
- Refresh Moli layout during snapshot waits
  lanyue-llk
  committed
  cd8128f
- Bound Moli dynamic layout preparation
  lanyue-llk
  committed
  7e97aaf
- Harden Moli layout actionability and deadlines
  lanyue-llk
  committed
  b68bcef
- Wait for snapshot document before Moli layout
  lanyue-llk
  committed
  85192f5
- Preserve native non-click actionability
  lanyue-llk
  committed
  a27cd84
- Preserve native targeted press actionability
  lanyue-llk
  committed
  df0c3c4
- Publish Moli layout without screenshot paint
  lanyue-llk
  committed
  5f14352
- Pin verified evaluator and repair route trajectories
  lanyue-llk
  committed
  26fda9c
- Capture same URL navigation evidence
  lanyue-llk
  committed
  57c56ac
- Complete Moli WebArena replay acceptance
  lanyue-llk
  committed
  2970532
- Add native CDP replay adapters
  lanyue-llk
  committed
  b513142
- Support single-target replay sessions
  lanyue-llk
  committed
  b5744ac
- Bootstrap Postmill in remote browser sessions
  lanyue-llk
  committed
  a5c16bd
- Compile catalog evidence for comparison reports
  lanyue-llk
  committed
  9bfb290
- Parameterize trajectory report inputs
  lanyue-llk
  committed
  e99bd12
- Generalize formal replay release scope
  lanyue-llk
  committed
  06281c3
- Separate browser session setup failures
  lanyue-llk
  committed
  5cee615
- Record replay failure phases explicitly
  lanyue-llk
  committed
  2804bc8
- Publish stable replay error excerpts
  lanyue-llk
  committed
  98acbc1
- Scope replay state reset to task sites
  lanyue-llk
  committed
  ffccb70
- Recover replay setup after browser exit
  lanyue-llk
  committed
  4b128c6
- Register three missing catalog entry points in naming gate
  
  authored and
  lanyue-llk
  committed
  009f61e
- Preserve web routes in report sanitization
  lanyue-llk
  committed
  adde745
- Avoid secret false positives in report URLs
  lanyue-llk
  committed
  8868448
- Harden replay report publication
  lanyue-llk
  committed
  8ef729e
- Handle partial CDP state reset support
  lanyue-llk
  committed
  aef81c7
- Retry storage cleanup after navigation
  lanyue-llk
  committed
  0019759
- Use fresh contexts for replay isolation
  lanyue-llk
  committed
  0df15e0
- Bound replay session setup time
  lanyue-llk
  committed
  4afeb25
- Isolate Obscura replay by browser process
  lanyue-llk
  committed
  5e6acbc
- Apply benchmark viewport to Obscura default context
  lanyue-llk
  committed
  8660d36
- Set Obscura viewport through native CDP
  lanyue-llk
  committed
  c6cb17d
- Harden Obscura replay and compact evidence
  lanyue-llk
  committed
  8baed29
- Add Kitesurf hosted replay runtime
  lanyue-llk
  committed
  13405ea
- Bridge Kitesurf requests to private benchmark sites
  lanyue-llk
  committed
  5d950a1
- Preserve Kitesurf bridge browser semantics
  lanyue-llk
  committed
  7073b2f
- Observe Kitesurf documents without locator internals
  lanyue-llk
  committed
  a3e5f32
- Preserve Kitesurf mutation bodies in replay evidence
  lanyue-llk
  committed
  021f528
- Preserve Kitesurf form submissions in replay evidence
  lanyue-llk
  committed
  cab5dfa
- Honor deployment host routes in Kitesurf replay
  lanyue-llk
  committed
  68ed631
- Keep Kitesurf fetch bodies inside the local bridge
  lanyue-llk
  committed
  2f4bb18
- Bind the current four-browser replay release
  lanyue-llk
  committed
  b6dc1b3
- Pin corrected WebArena evaluator revision
  lanyue-llk
  committed
  2ffa720
- support evaluator transitions across replay revisions
  lanyue-llk
  committed
  819e82d
- render browser and method counts separately
  lanyue-llk
  committed
  1ae99ca
- enforce browser and method populations
  lanyue-llk
  committed
  8b58885
- Bind replay transitions to exact evidence
  lanyue-llk
  committed
  fad90a6
- Rebind WebArena report dataset
  lanyue-llk
  committed
  f2e2dad
- Generalize replay release diagnostics
  lanyue-llk
  committed
  b572470
- Publish six-method WebArena replay comparison
  lanyue-llk
  committed
  768b581
- Clarify replay backend and improve report tables
  lanyue-llk
  committed
  8e5ba8d
- Bind corrected trajectory replay report
  lanyue-llk
  committed
  7cbbeb1
- Improve replay report failure storytelling
  lanyue-llk
  committed
  f2201ec
- Bind replay evidence and identify repaired builds
  lanyue-llk
  committed
  8aac1d3
- fix(trajectory-replay): close hosted execution boundaries
  lanyue-llk
  committed
  5c2b0fd
- fix(trajectory-replay): derive failure copy from participants
  lanyue-llk
  committed
  9a51da6

Commits

Commits on Sep 29, 2026