{ "title": "Addition conditioning: anchor frame vs captions", "ckpt": "text_both_audiofix_scratch_0811 @ epoch028-global_step112500", "split": "test_first50_fixed.jsonl (first 10 with all arms shown)", "grid": { "size": 256, "fps": 16, "frames": 81, "sr": 16000, "duration_s": 5.0625 }, "arms": { "bare": { "label": "bare noun", "anchor": false, "prompt": "object name only" }, "withref": { "label": "+ first frame", "anchor": true, "prompt": "object name only" }, "short": { "label": "short caption", "anchor": false, "prompt": "3-8 words" }, "long": { "label": "long caption", "anchor": false, "prompt": "8-18 words" }, "refshort": { "label": "first frame + short", "anchor": true, "prompt": "3-8 words" }, "reflong": { "label": "first frame + long", "anchor": true, "prompt": "8-18 words" } }, "aggregate": { "withref": { "n": 50, "gap": 2.03, "median": 2.25, "moved": "18/50" }, "refshort": { "n": 50, "gap": 2.13, "median": 1.9, "moved": "18/50" }, "reflong": { "n": 50, "gap": 4.02, "median": 4.05, "moved": "13/50" }, "bare": { "n": 50, "gap": 2.64, "median": 2.85, "moved": "17/50" }, "short": { "n": 50, "gap": 3.37, "median": 4.1, "moved": "16/50" }, "long": { "n": 50, "gap": 5.49, "median": 4.95, "moved": "10/50" } }, "entries": [ { "video_id": "NU2oWhlEWss", "object": "vehicles", "caption_short": "White and black articulated bus", "caption_long": "white and black articulated bus with 'COLT45' advertisement, moving slowly on a road.", "clips": { "input": "videos/NU2oWhlEWss_input.mp4", "gt": "videos/NU2oWhlEWss_gt.mp4", "anchor": "videos/NU2oWhlEWss_anchor.jpg", "withref": "videos/NU2oWhlEWss_withref.mp4", "refshort": "videos/NU2oWhlEWss_refshort.mp4", "reflong": "videos/NU2oWhlEWss_reflong.mp4", "bare": "videos/NU2oWhlEWss_bare.mp4", "short": "videos/NU2oWhlEWss_short.mp4", "long": "videos/NU2oWhlEWss_long.mp4" }, "metrics": { "withref": { "v_gap": -3.93, "a_gap": -1.1 }, "refshort": { "v_gap": -3.13, "a_gap": -1.09 }, "reflong": { "v_gap": -1.95, "a_gap": -1.15 }, "bare": { "v_gap": -5.54, "a_gap": -1.09 }, "short": { "v_gap": -4.24, "a_gap": -1.09 }, "long": { "v_gap": 17.2, "a_gap": -1.15 } } }, { "video_id": "X7t8wKNIAP4", "object": "birds", "caption_short": "Two gray pigeons with black wing bars", "caption_long": "Two gray pigeons with black wing bars stand close together, one facing forward, the other turned slightly.", "clips": { "input": "videos/X7t8wKNIAP4_input.mp4", "gt": "videos/X7t8wKNIAP4_gt.mp4", "anchor": "videos/X7t8wKNIAP4_anchor.jpg", "withref": "videos/X7t8wKNIAP4_withref.mp4", "refshort": "videos/X7t8wKNIAP4_refshort.mp4", "reflong": "videos/X7t8wKNIAP4_reflong.mp4", "bare": "videos/X7t8wKNIAP4_bare.mp4", "short": "videos/X7t8wKNIAP4_short.mp4", "long": "videos/X7t8wKNIAP4_long.mp4" }, "metrics": { "withref": { "v_gap": 6.44, "a_gap": -1.28 }, "refshort": { "v_gap": 7.42, "a_gap": -1.23 }, "reflong": { "v_gap": 10.58, "a_gap": -1.7 }, "bare": { "v_gap": 5.63, "a_gap": -1.31 }, "short": { "v_gap": 6.29, "a_gap": -1.26 }, "long": { "v_gap": 10.87, "a_gap": -1.71 } } }, { "video_id": "Rumslm1EQhU", "object": "water", "caption_short": "clear water in a toilet bowl", "caption_long": "Clear water sits still in the white ceramic toilet bowl, reflecting light from above.", "clips": { "input": "videos/Rumslm1EQhU_input.mp4", "gt": "videos/Rumslm1EQhU_gt.mp4", "anchor": "videos/Rumslm1EQhU_anchor.jpg", "withref": "videos/Rumslm1EQhU_withref.mp4", "refshort": "videos/Rumslm1EQhU_refshort.mp4", "reflong": "videos/Rumslm1EQhU_reflong.mp4", "bare": "videos/Rumslm1EQhU_bare.mp4", "short": "videos/Rumslm1EQhU_short.mp4", "long": "videos/Rumslm1EQhU_long.mp4" }, "metrics": { "withref": { "v_gap": -24.05, "a_gap": 0.61 }, "refshort": { "v_gap": -24.08, "a_gap": 0.6 }, "reflong": { "v_gap": -20.52, "a_gap": 0.55 }, "bare": { "v_gap": -22.8, "a_gap": 0.62 }, "short": { "v_gap": -22.33, "a_gap": 0.6 }, "long": { "v_gap": -16.61, "a_gap": 0.56 } } }, { "video_id": "65vCgp9aKGU_000116", "object": "birds", "caption_short": "white cockatiels with yellow crests", "caption_long": "White cockatiels with yellow crests perched on wooden bars, some moving slightly.", "clips": { "input": "videos/65vCgp9aKGU_000116_input.mp4", "gt": "videos/65vCgp9aKGU_000116_gt.mp4", "anchor": "videos/65vCgp9aKGU_000116_anchor.jpg", "withref": "videos/65vCgp9aKGU_000116_withref.mp4", "refshort": "videos/65vCgp9aKGU_000116_refshort.mp4", "reflong": "videos/65vCgp9aKGU_000116_reflong.mp4", "bare": "videos/65vCgp9aKGU_000116_bare.mp4", "short": "videos/65vCgp9aKGU_000116_short.mp4", "long": "videos/65vCgp9aKGU_000116_long.mp4" }, "metrics": { "withref": { "v_gap": 11.89, "a_gap": 0.63 }, "refshort": { "v_gap": 12.05, "a_gap": 0.85 }, "reflong": { "v_gap": 13.88, "a_gap": 1.04 }, "bare": { "v_gap": 10.27, "a_gap": 0.64 }, "short": { "v_gap": 10.42, "a_gap": 0.86 }, "long": { "v_gap": 11.89, "a_gap": 1.03 } } }, { "video_id": "hVnvazN3U0s_000060", "object": "motorboat engine", "caption_short": "No visible motorboat engine", "caption_long": "No visible motorboat engine in the provided frames.", "clips": { "input": "videos/hVnvazN3U0s_000060_input.mp4", "gt": "videos/hVnvazN3U0s_000060_gt.mp4", "anchor": "videos/hVnvazN3U0s_000060_anchor.jpg", "withref": "videos/hVnvazN3U0s_000060_withref.mp4", "refshort": "videos/hVnvazN3U0s_000060_refshort.mp4", "reflong": "videos/hVnvazN3U0s_000060_reflong.mp4", "bare": "videos/hVnvazN3U0s_000060_bare.mp4", "short": "videos/hVnvazN3U0s_000060_short.mp4", "long": "videos/hVnvazN3U0s_000060_long.mp4" }, "metrics": { "withref": { "v_gap": -12.22, "a_gap": -0.24 }, "refshort": { "v_gap": -10.76, "a_gap": -0.02 }, "reflong": { "v_gap": 12.78, "a_gap": -0.0 }, "bare": { "v_gap": -11.12, "a_gap": -0.15 }, "short": { "v_gap": -9.92, "a_gap": 0.09 }, "long": { "v_gap": -8.41, "a_gap": 0.1 } } }, { "video_id": "u_8JdjQOo-o_000151", "object": "bagpipe", "caption_short": "black bagpipe with tassels", "caption_long": "Black bagpipe with tassels, held by player, bag inflated, drones and chanter visible.", "clips": { "input": "videos/u_8JdjQOo-o_000151_input.mp4", "gt": "videos/u_8JdjQOo-o_000151_gt.mp4", "anchor": "videos/u_8JdjQOo-o_000151_anchor.jpg", "withref": "videos/u_8JdjQOo-o_000151_withref.mp4", "refshort": "videos/u_8JdjQOo-o_000151_refshort.mp4", "reflong": "videos/u_8JdjQOo-o_000151_reflong.mp4", "bare": "videos/u_8JdjQOo-o_000151_bare.mp4", "short": "videos/u_8JdjQOo-o_000151_short.mp4", "long": "videos/u_8JdjQOo-o_000151_long.mp4" }, "metrics": { "withref": { "v_gap": 6.7, "a_gap": -0.28 }, "refshort": { "v_gap": 6.82, "a_gap": -1.08 }, "reflong": { "v_gap": 7.41, "a_gap": -0.27 }, "bare": { "v_gap": 5.77, "a_gap": -0.3 }, "short": { "v_gap": 5.86, "a_gap": -1.04 }, "long": { "v_gap": 6.64, "a_gap": -0.12 } } }, { "video_id": "TbTUbDC-IuQ_000249", "object": "bass line", "caption_short": "shiny silver steel drum", "caption_long": "large, shiny silver steel drum being played with purple-tipped mallets, producing resonant bass tones.", "clips": { "input": "videos/TbTUbDC-IuQ_000249_input.mp4", "gt": "videos/TbTUbDC-IuQ_000249_gt.mp4", "anchor": "videos/TbTUbDC-IuQ_000249_anchor.jpg", "withref": "videos/TbTUbDC-IuQ_000249_withref.mp4", "refshort": "videos/TbTUbDC-IuQ_000249_refshort.mp4", "reflong": "videos/TbTUbDC-IuQ_000249_reflong.mp4", "bare": "videos/TbTUbDC-IuQ_000249_bare.mp4", "short": "videos/TbTUbDC-IuQ_000249_short.mp4", "long": "videos/TbTUbDC-IuQ_000249_long.mp4" }, "metrics": { "withref": { "v_gap": 10.89, "a_gap": 0.16 }, "refshort": { "v_gap": 13.58, "a_gap": 0.58 }, "reflong": { "v_gap": 16.45, "a_gap": 0.49 }, "bare": { "v_gap": 3.88, "a_gap": 0.15 }, "short": { "v_gap": 7.32, "a_gap": 0.6 }, "long": { "v_gap": 11.12, "a_gap": 0.48 } } }, { "video_id": "GwA7hMcmJ6Q", "object": "piano", "caption_short": "black grand piano with glossy finish", "caption_long": "black grand piano with a glossy finish, reflecting light on its smooth surface.", "clips": { "input": "videos/GwA7hMcmJ6Q_input.mp4", "gt": "videos/GwA7hMcmJ6Q_gt.mp4", "anchor": "videos/GwA7hMcmJ6Q_anchor.jpg", "withref": "videos/GwA7hMcmJ6Q_withref.mp4", "refshort": "videos/GwA7hMcmJ6Q_refshort.mp4", "reflong": "videos/GwA7hMcmJ6Q_reflong.mp4", "bare": "videos/GwA7hMcmJ6Q_bare.mp4", "short": "videos/GwA7hMcmJ6Q_short.mp4", "long": "videos/GwA7hMcmJ6Q_long.mp4" }, "metrics": { "withref": { "v_gap": 0.26, "a_gap": -0.29 }, "refshort": { "v_gap": 0.16, "a_gap": -0.58 }, "reflong": { "v_gap": 0.07, "a_gap": -0.18 }, "bare": { "v_gap": 2.93, "a_gap": -0.26 }, "short": { "v_gap": 2.95, "a_gap": -0.55 }, "long": { "v_gap": 3.39, "a_gap": -0.22 } } }, { "video_id": "884QLkT8sBQ_000098", "object": "musician", "caption_short": "Young boy playing tabla drums", "caption_long": "young boy in white shirt plays tabla drums with focused expression and precise hand movements.", "clips": { "input": "videos/884QLkT8sBQ_000098_input.mp4", "gt": "videos/884QLkT8sBQ_000098_gt.mp4", "anchor": "videos/884QLkT8sBQ_000098_anchor.jpg", "withref": "videos/884QLkT8sBQ_000098_withref.mp4", "refshort": "videos/884QLkT8sBQ_000098_refshort.mp4", "reflong": "videos/884QLkT8sBQ_000098_reflong.mp4", "bare": "videos/884QLkT8sBQ_000098_bare.mp4", "short": "videos/884QLkT8sBQ_000098_short.mp4", "long": "videos/884QLkT8sBQ_000098_long.mp4" }, "metrics": { "withref": { "v_gap": 2.15, "a_gap": -0.35 }, "refshort": { "v_gap": 1.12, "a_gap": -0.67 }, "reflong": { "v_gap": 3.87, "a_gap": -0.66 }, "bare": { "v_gap": 2.32, "a_gap": -0.49 }, "short": { "v_gap": 6.46, "a_gap": -0.78 }, "long": { "v_gap": 6.73, "a_gap": -0.83 } } }, { "video_id": "NhWaVKzBX_g_000056", "object": "powerful engine", "caption_short": "Tall slender rocket engine", "caption_long": "tall, slender rocket engine with a dark body and bright exhaust trail propels upward.", "clips": { "input": "videos/NhWaVKzBX_g_000056_input.mp4", "gt": "videos/NhWaVKzBX_g_000056_gt.mp4", "anchor": "videos/NhWaVKzBX_g_000056_anchor.jpg", "withref": "videos/NhWaVKzBX_g_000056_withref.mp4", "refshort": "videos/NhWaVKzBX_g_000056_refshort.mp4", "reflong": "videos/NhWaVKzBX_g_000056_reflong.mp4", "bare": "videos/NhWaVKzBX_g_000056_bare.mp4", "short": "videos/NhWaVKzBX_g_000056_short.mp4", "long": "videos/NhWaVKzBX_g_000056_long.mp4" }, "metrics": { "withref": { "v_gap": 9.48, "a_gap": -0.08 }, "refshort": { "v_gap": 9.67, "a_gap": -0.14 }, "reflong": { "v_gap": 9.77, "a_gap": -0.1 }, "bare": { "v_gap": 9.03, "a_gap": 0.11 }, "short": { "v_gap": 8.92, "a_gap": -0.03 }, "long": { "v_gap": 8.87, "a_gap": -0.1 } } } ] }