From 58ca0d50084ef62d28aba23e4196ffb01ddb7a52 Mon Sep 17 00:00:00 2001 From: pi Date: Sun, 30 Aug 2026 04:18:37 +1200 Subject: [PATCH] Fix grammar: subject_verb_agreement_modalities --- mmeb_v3_eval/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/mmeb_v3_eval/README.md b/mmeb_v3_eval/README.md index 3528194..d3a9d16 100644 --- a/mmeb_v3_eval/README.md +++ b/mmeb_v3_eval/README.md @@ -107,7 +107,7 @@ datasets are skipped on re-run (embedding pickles are reused). ## Notes & limitations - **Audio tasks are not supported** (WeMM-Embedding has no audio tower). The - default `MODALITIES` in the run scripts therefore excludes `audio`. + default `MODALITIES` in the run scripts therefore exclude `audio`. - `` token: appended automatically by the model-side chat template / tokenizer post-processor — do not insert it manually. - Video inputs follow the WeMM frame-bundle convention: `video_grid_thw` is