Sep 25 19 days ago 22:01 Rename assets/JevBench Results.png to assets/JevBench-results.png 22:01 Update README.md 22:00 Upload 2 files 21:59 Delete assets/svgviewer-png-output.png 21:59 Delete assets/svgviewer-output.svg 21:54 Delete assets/jevbench-m3-results.png 21:53 Delete assets/jevbench-m3-results.svg 21:53 Update README.md 21:51 Upload 2 files 21:03 Restore scoped JevBench comparison in updated visual style 20:43 Clarify JEVision evaluation claims and update graphics 20:23 Refresh JEVision model card architecture hero 19:58 Restore prior JEVision files and card graphics 18:50 Present JEVision architecture and restore verified examples 18:10 Add JEVision capability comparison graphic Adds an evidence-aware comparison of Jev, KEV-0.8B and JEVision, including context limits, image routing and text-only JevBench results. The README clarifies in-sample scoring and long-context scope. 17:41 Add JevBench results chart and JEVision artwork Show the benchmark-tuned public-panel result separately from grouped holdout performance. Include generated brand artwork and disclose that raw Qwen3.5 was not independently scored. 16:06 Add recorded 77K text acceptance summary 14:45 Add runnable 75K+ text-context example 14:00 Add an attributed real-photo image example 12:41 Bundle JEVision without benchmark scores Sep 24 20 days ago 08:59 Document JEVision training data 08:04 Add JEVision model card 07:22 Publish JEVision POC assets 07:21 Publish JEVision POC assets 07:21 Publish JEVision POC assets 07:21 Publish JEVision POC assets 07:21 Publish JEVision POC assets 07:21 Publish JEVision POC assets 07:21 initial commit 
