Delete svg/llava-1.5-7b-hf_cache.json

Update reference_with_cvbench.xlsx with latest CVBench scores

Update svg/InternVL3-8B-Instruct_cache.json (v19conv SVG-500 in-flight, all-fixes run, 0% empty)

Update svg/Qwen2.5-VL-7B-Instruct_cache.json (v19conv SVG-500 in-flight, all-fixes run, 0% empty)

Update svg/llava-1.5-7b-hf_cache.json (v19conv SVG-500 in-flight, all-fixes run, 0% empty)

deepeyes tab: fill results from eval/results/{deepeyes,input_ori_image,deepeyesv1v8_thinklite_rotation-disabled_ori_image_input}

Upload demistifyTWI-results.xlsx

Add v19ban2type/gemma-3-12b-it.json (v19ban2type VG-100 result)

Add v19ban2type/Qwen2.5-VL-7B-Instruct.json (v19ban2type VG-100 result)

Add v19ban2type/llava-1.5-7b-hf.json (v19ban2type VG-100 result)

Delete Qwen cache - run was on pre-fix code (race + no reset-forwarding)

Update reference_with_cvbench.xlsx with latest CVBench scores

Update reference_with_cvbench.xlsx with latest CVBench scores

Delete llava cache - was single-shot (no multi-turn wrapper); fixed in v19conv branch

Delete svg/InternVL3-8B-Instruct_cache.json (InternVL image-bug-impacted v19conv cache)

Delete svg/InternVL2_5-8B_cache.json (InternVL image-bug-impacted v19conv cache)

Delete svg/InternVL2-8B_cache.json (InternVL image-bug-impacted v19conv cache)

update svg/Qwen2.5-VL-7B-Instruct_cache.json (v19conv SVG-500 partial)

update svg/llava-1.5-7b-hf_cache.json (v19conv SVG-500 partial)

Upload partial v19conv SVG-500 results (in-flight + stopped runs)

Add gemini-2.5-pro judge_encode results (5549 samples)

Add gpt-5.4-2026-03-05 judge_encode results (5549 samples)

Upload encoded_v2.tar (~19 GB single archive)

Upload docs/

Add files using upload-large-folder tool

Initial upload of participant attributes dataset

Delete bert-base-uncased.zip

Upload bert-base-uncased.zip

initial commit