WEBVTT

1
00:00:00.260 --> 00:00:07.600
Imagine an original animation: a blue ball should move smoothly
right, keeping its appearance and continuous motion.
想象原创动画：蓝球应平滑向右移动，不改变外观，也不突然跳跃。

2
00:00:08.500 --> 00:00:14.120
One version becomes blurry, making the ball hard to inspect even
while its motion follows the instruction.
一个版本变模糊，难以看清球，但运动仍符合指令。

3
00:00:14.940 --> 00:00:21.720
Another looks crisp but shows a red ball, contradicting the
requested color despite an otherwise plausible movement.
另一个画面清晰却显示红球，虽然运动合理，仍违背所需颜色。

4
00:00:22.680 --> 00:00:28.380
A third has the right color but an unexplained jump, creating a
different kind of problem over time.
第三个颜色正确却无故跳跃，形成另一种随时间变化的问题。

5
00:00:29.280 --> 00:00:35.480
These are original teaching animations, not research samples,
measured outputs, or copied paper videos.
这些动画是原创教学示例，并非研究样本、实测输出或复制论文视频。

6
00:00:36.260 --> 00:00:42.340
Visual quality, prompt alignment, and physical consistency
deserve separate questions before a score.
重点是区别：评分前，视觉质量、指令对齐和物理一致性应分别考察。

7
00:00:42.913 --> 00:00:50.533
Visual quality concerns clarity, smoothness, stable brightness,
distortions, and other visible viewing defects.
视觉质量考察清晰度、平滑程度、亮度稳定、畸变等影响观看的缺陷。

8
00:00:51.693 --> 00:00:59.033
Text alignment checks subjects, actions, details, style, and
event order against the generation prompt.
文本对齐把主体、动作、细节、风格和事件顺序与生成提示比较。

9
00:01:00.213 --> 00:01:06.553
Physical consistency checks everyday expectations and plausible
object relationships as the video unfolds.
物理一致性检查行为是否符合日常预期及随时间展开的合理物体关系。

10
00:01:07.733 --> 00:01:14.453
Temporal consistency belongs to visual quality; physical and
common-sense consistency form a separate dimension.
论文把时间一致性纳入视觉质量，另设物理与常识一致性维度。

11
00:01:15.633 --> 00:01:21.533
These scores describe different aspects of one video, rather
than three interchangeable overall judgments.
三个得分因此描述同一候选的不同方面，而不是整体判断的三个可互换版本。

12
00:01:22.433 --> 00:01:29.493
A defect can affect several axes, so separate scores expose
different concerns rather than disjoint interventions.
缺陷可影响多个轴；分开评分展示不同问题，而非完全独立的干预。

13
00:01:30.120 --> 00:01:37.220
VideoFeedback two pairs generated videos with dimension scores
and diagnostic explanations for training the critic.
VideoFeedback2 将生成视频与维度得分和解释配对，提供评价和诊断语言样本。

14
00:01:38.040 --> 00:01:44.360
The dataset contains twenty-seven thousand one hundred
sixty-eight videos, evaluated along all three axes.
处理后包含二万七千一百六十八视频，每例沿三个轴评估。

15
00:01:44.960 --> 00:01:51.380
It contains eighty-one thousand five hundred four dimension
scores with rationales, rather than one rating per video.
它包含八万一千五百零四个带理由维度得分，而非每视频一个独立评级。

16
00:01:52.160 --> 00:01:59.360
Prompts include multiple actions, readable text, and camera
motion; videos span diverse systems and quality levels.
提示包含多动作、可读文字、相机运动；视频覆盖多系统与不同质量水平。

17
00:02:00.280 --> 00:02:06.480
Five hundred held-out videos form VideoScore Bench version two,
separating the in-domain test from training.
留出五百视频组成 VideoScore-Bench-v2，形成与训练分开的域内测试。

18
00:02:07.420 --> 00:02:13.060
These examples teach what to inspect and how to describe
defects, beyond supplying numerical labels.
这教授评价器检查内容及描述观察问题的方式，而不是只提供数值标签。

19
00:02:13.727 --> 00:02:20.367
Human annotators assign scores from one to five and write short
diagnostic comments for each of the three dimensions.
人工标注者给三个维度各赋一至五分，并写简短诊断评论。

20
00:02:21.207 --> 00:02:28.107
An LLM sees frames, instructions, and comments without human
scores, then generates a detailed assessment.
语言模型读取采样帧、指令及不含人工得分的评论，生成更详细评价。

21
00:02:28.967 --> 00:02:35.187
Scores are reconciled with explicit difference thresholds,
including rescoring when human and model disagree.
预测得分按明确差异门槛与人工评价协调，差异大时重新评分。

22
00:02:36.087 --> 00:02:42.787
A later step aligns rationale wording with final scores,
preserving assessment meaning while resolving inconsistencies.
后续对齐调整理由措辞，使其匹配最终得分，保留评价含义并消除描述不一致。

23
00:02:43.787 --> 00:02:50.107
Detailed training rationales use LLM-assisted expansions; they
do not all come directly from human annotation.
详细训练理由使用语言模型辅助扩展，并非全部直接来自人工标注。

24
00:02:50.847 --> 00:02:57.427
The critic describes observed evidence before giving scores,
making the explanation inspectable alongside its output.
评价器学习先描述观察证据再评分，让理由与三轴输出一起可供检查。

25
00:02:58.013 --> 00:03:05.813
Training begins with Qwen two point five VL seven B Instruct and
a cold-start stage of supervised fine-tuning.
训练从 Qwen2.5-VL-7B-Instruct 开始，先进行监督微调冷启动。

26
00:03:06.693 --> 00:03:12.653
Cold-start training teaches the task and output structure:
analysis followed by three dimension scores.
冷启动训练教授任务与输出结构：先分析，再给出三个维度得分。

27
00:03:13.613 --> 00:03:18.873
GRPO refines that checkpoint using rewards for agreement with
the target dimension scores.
随后 GRPO 强化学习改进监督检查点，奖励依据与目标维度得分的一致性。

28
00:03:19.813 --> 00:03:27.233
Nonzero accuracy reward requires every dimension within one
reference point; exact agreement receives stronger reward.
非零准确奖励要求每维度与参考相差不超过一点，精确一致获得更强奖励。

29
00:03:28.173 --> 00:03:35.013
Reported results use the three-hundred-step RL checkpoint, with
video frames sampled at two per second during inference.
报告结果使用三百步强化学习检查点，推理每秒采样两帧。

30
00:03:35.893 --> 00:03:42.053
This trains a video critic, rather than simultaneously
optimizing the generator that produced its candidate inputs.
这训练评价生成候选的评价器，不同时优化产生候选的生成器。

31
00:03:42.700 --> 00:03:49.400
Exact point-score accuracy averages the visual-quality,
alignment, and physics judgments on the in-domain test.
域内测试精确评分准确率在视觉质量、文本对齐和物理一致性上平均。

32
00:03:50.220 --> 00:03:57.260
VideoScore one point one scores thirty-eight point four one;
VideoScore two reaches forty-four point three five.
VideoScore1.1 为三十八点四一，VideoScore2 达到四十四点三五。

33
00:03:58.120 --> 00:04:04.040
Relaxed accuracy uses a different tolerance, counting
predictions within one point of the reference score as correct.
宽松准确率使用不同容差，把与参考相差一点以内的预测计为正确。

34
00:04:04.760 --> 00:04:09.680
Relaxed accuracy reaches ninety point seven eight, a different
event from exact score matching.
VideoScore2 在宽松指标下为九十点七八，这与精确匹配是不同事件。

35
00:04:10.680 --> 00:04:16.580
Correlation measures how predicted and reference scores vary
together, rather than matching individual integer ratings.
相关性是另一报告指标，描述预测与参考如何共同变化，而非匹配单个整数评级。

36
00:04:17.420 --> 00:04:23.260
Keeping metrics separate prevents a high tolerance-based score
from being mistaken for equally high exact accuracy.
区分这些指标，避免把高容差得分解释为同样高的精确判断比例。

37
00:04:23.880 --> 00:04:30.200
Four out-of-domain benchmarks include two preference settings
and two point-score settings with distinct targets.
域外比较覆盖四基准，包括两个成对偏好和两个评分设置，评价目标不同。

38
00:04:30.980 --> 00:04:37.580
Preference tests compare videos in pairs; point-score tests
compare numerical judgments with human references.
偏好测试选择一对视频中更好的；评分测试将数值评价与人工参考比较。

39
00:04:38.480 --> 00:04:44.100
Dimension mapping and score rescaling allow these distinct
benchmark results to be summarized together.
维度经过映射、得分经过缩放，让论文汇总这些不同基准设置的性能。

40
00:04:45.080 --> 00:04:52.660
Table six reports fifty point three seven for VideoScore two,
against forty-six point zero five for Q-Insight.
表六报告 VideoScore2 五十点三七，Q-Insight 四十六点零五。

41
00:04:53.400 --> 00:05:00.680
This composite mixes benchmark metrics; it is neither universal
accuracy nor victory on every individual benchmark.
这是混合基准指标的综合均值，不是通用视频判断准确率，也不是每基准获胜。

42
00:05:01.500 --> 00:05:07.440
The scoped comparison supports transfer of three-axis assessment
under the reported benchmark mappings.
它支持三轴评价在报告映射下的迁移，同时保留各比较的具体任务范围。

43
00:05:07.980 --> 00:05:14.380
A trained critic compares generated alternatives, using its
assessments to choose a more promising supplied candidate.
训练后的评价器比较生成视频备选项，通过评价从输入候选中选择更有希望的输出。

44
00:05:15.220 --> 00:05:21.780
Best-of-five selection uses five hundred prompts and six
moderate or poorer generators, compared with random choice.
论文在五百提示和六个中等或较差生成器上测试五选一，并与随机选择比较。

45
00:05:22.680 --> 00:05:28.920
Selection chooses an existing output, rather than changing
generator parameters while deciding which video to return.
选择针对已有输出，并不在决定返回哪个视频时改变生成器参数。

46
00:05:29.820 --> 00:05:36.080
For the blue ball, clarity, requested content, and plausible
motion remain different candidate-comparison questions.
回到蓝球：比较备选项时，清晰度、所需内容与合理运动仍是不同问题。

47
00:05:36.940 --> 00:05:42.400
A reasoned critic identifies its observations, their dimension,
and how the evidence supports a score.
带理由评价器解释观察到了什么、影响哪个维度，以及证据怎样支持最终得分。

48
00:05:43.340 --> 00:05:49.400
Thinking before scoring makes judgments inspectable, while
reliability remains tied to the measured tasks and setups.
先思考再评分让判断更可检查，而实测可靠性仍限定在评估任务与配置。