公开语音AI基准测试越来越显示出模型已达到人类水平的表现。然而,这些分数并不总能反映模型在现实世界中的实际表现。由于公开基准测试是开放的且被广泛使用,模型也可能针对测试本身进行优化。它们的分数提升可能是因为学到了基准测试特有的模式,而不是因为底层任务能力变得更强。原因之一是传统基准测试忽略了许多使语音系统在实践中的可靠性、自然度、上下文适配性和有效性得以保障的条件与品质。正因如此,我们最近在Real World VoiceEQ、Open-ASR排行榜和远场ASR排行榜中引入了保留集(held-out sets),以衡量更多在现实使用中真正重要的指标。
然而,仅靠更广泛的评测并不能解决这个问题。这一现象有时被称为基准优化或“benchmaxxing”,在机器学习领域经常被讨论,但在语音识别中一直难以量化衡量。
我们最新的研究引入了三项测试来帮助量化这一现象。我们评估了11个广泛使用的开源ASR模型,发现其中几个得分最高的系统会复现VoxPopuli英语和LibriSpeech(clean、other)数据集中的基准转录文本——即使音频内容与之矛盾、相关词语已被静音,或音频同样支持两种不同的书面形式时也是如此。
在某些情况下,模型似乎不仅依赖所说的内容,还依赖表明它们正在接受哪个基准测试的细微声学线索。因此,它们的分数高估了它们在更广泛场景下转录语音的能力。
参考文本不一致(VoxPopuli案例研究)
众所周知,VoxPopuli包含大量转录错误(这也是Artificial Analysis发布清理版本的原因)。我们的共识不一致性探针测试了当领先的ASR模型遇到这些错误时会发生什么:它们是准确转录音频实际所说的内容,还是复现基准测试中错误的参考转录文本?
为了在大规模范围内测试这一点,我们使用了一个由独立模型组成的集成,这些模型因其较低的音素错误率(PER)而被选中。PER 衡量书面转写与音频中声音的匹配程度,因此可以作为模型转写所听内容的忠实度的有用指标。集成结果可用于标记那些模型一致不同意基准参考转写的案例。然后,我们将这些被标记案例中的一部分样本与人工标注进行比较,以验证修正后的转写。
例如,一段 VoxPopuli 片段中明显包含“Thank you, Mr. President”这句话,但参考转写却省略了“Thank you”。在我们测试的 11 个模型中,有 6 个复现了基准的错误转写——即使这与音频相矛盾,也给出了“预期”答案。在真实片段上,格式遵循相同的模式:省略“Thank you”的模型也复现了基准的标点风格,将“Mr”写成不带句号的形式,而包含该可听短语的模型则倾向于将“Mr.”写成带句号的形式。
当我们使用来自欧盟议会录音或通用语音的新采集声音呈现相同内容时,这种行为往往会减弱或消失。在下面的样本中,除了一个模型外,所有模型都转而针对新议会录音的克隆版本转写出忠实于音频的文本。这表明模型正在响应声学线索,这些线索帮助它们识别基准成员身份,从而即使与音频相矛盾,也会生成预期的转写。
此片段对应的参考转写文本为“主席先生,我对这一程序还有一项不满,即它并非秘密进行。”下方三段音频中的实际语音内容完全相同,且前面都带有一声可听见的“谢谢”——这些克隆音频是对那句真实句子的文本转语音演绎,因此三者的礼貌用语都能被听见。绿色高亮和✅标记表示转写结果包含了可听见的“谢谢”;红色高亮和❌标记表示转写结果复现了基准测试中的错误遗漏。所有转写均为模型原始输出,未经任何归一化处理——大小写和标点完全按生成结果保留,包括某些模型输出的小写形式。
原始 VoxPopuli 录音
同一说话人的语音克隆
在每款模型训练截止日期之后录制的议会发言人克隆
| 模型 | 真实片段 | 同说话人克隆 | ep-fresh 克隆 |
|---|---|---|---|
| CohereLabs/cohere-transcribe-03-2026 | ❌ 主席先生…… | ❌ 主席先生…… | ✅ 谢谢,主席先生…… |
| nvidia/canary-qwen-2.5b | ❌ 主席先生…… | ❌ 主席先生…… | ✅ 谢谢,主席先生…… |
| ibm-granite/granite-speech-4.1-2b | ❌ 主席先生…… | ❌ 主席先生…… | ✅ 谢谢,主席先生…… |
| microsoft/Phi-4-multimodal-instruct | ❌ 主席先生…… | ❌ 主席先生…… | ❌ 主席先生…… |
| nvidia/parakeet-tdt-0.6b-v2 | ❌ 主席先生…… | ✅ 谢谢,主席先生…… | ✅ 谢谢,主席先生…… |
| bosonai/higgs-audio-v3-8b-stt-v2 | ❌ 主席先生…… | ❌ 主席先生…… | ✅ 谢谢,主席先生…… |
| Qwen/Qwen3-ASR-0.6B-hf | ✅ 谢谢,主席先生…… | ✅ 谢谢,主席先生…… | ✅ 谢谢,主席先生…… |
| mistralai/Voxtral-Mini-3B-2507 | ✅ 谢谢,主席先生…… | ✅ 谢谢,主席先生…… | ✅ 谢谢,主席先生…… |
| moonshotai/Kimi-Audio-7B-Instruct | ✅ 谢谢,主席先生…… | ✅ 谢谢,主席先生…… | ✅ 谢谢,主席先生…… |
| openai/whisper-large-v3 | ✅ 谢谢,主席先生…… | ✅ 谢谢,主席先生…… | ✅ 谢谢,主席先生…… |
| moonshine-ai/moonshine-streaming-medium | ✅ 谢谢,主席先生…… | ✅ 谢谢,主席先生…… | ✅ 谢谢,主席先生…… |
| 在 11 个片段中遗漏礼貌用语(❌)的数量 | 6 | 5 | 1 |
Parakeet 是唯一一个会在真实音频片段上复现基准错误、又在同说话人克隆音频上给出正确结果的模型。Phi-4 则是唯一一个在“全新说话人克隆”上仍然丢失礼貌用语的模型。当我们改用与任何议会录音无关的通用 TTS 音色重新合成该句子时,全部十一个模型都恢复了礼貌用语。
结果表明,这一问题既普遍又具有实质意义。我们的方法在分析的 VoxPopuli 测试片段中,标记出了 40% 的潜在参考转写错误,涉及约 3% 的全部参考词。
表现出“基准优化”行为的模型,有 18%–30% 的概率复现错误的参考转写。下方的散点图将 VoxPopuli 词错误率(WER)放在 x 轴,与各模型复现基准错误参考(而非共识修正)的频率进行对比。WER 最低——也就是报告基准性能最强——的模型,恰恰也是最容易复现这些错误的模型。

掩码实体检索
在共识分歧探测的基础上,我们刻意将测试数据集音频样本中的数字静音,然后让模型转写所听到的内容。数字在音频中根本不存在,因此模型不应输出任何数字,更不用说输出文本中的那个确切数字了。
其中一些数字具有一定可预测性(尽管模型仍不太可能预测出来),但另一些则相当出人意料。下面的片段结合了两种探测方式,既展示了模型如何复现参考转写错误(包括一个错误的数字),也展示了一个模型甚至在数字被静音的情况下自动补全了一个相对随机的年份(2011)。在下方每个模型对应的行中:
- 绿色高亮加删除线标记的是模型正确未复现的参考转写词(忠于音频);
- 绿色高亮加下划线标记的是模型在参考转写错误处做出的正确、忠于音频的插入。
- 红色高亮(纯文本)复现了参考转写文本中错误且无音频依据的内容:保留了“Mr President”,在音频实际说的是“one thousand six hundred”的地方写成了“more than 1 amendments”,补全了被静音的年份“2011”,并以“plenary”结尾。
2011 年预算草案(数字被遮蔽)
| 参考文本 | 主席先生,在预算委员会中,我们对 2011 年预算草案的 1 项以上修正案进行了表决……并在全体会议上进行了表决。 |
|---|---|
| 音频实际内容 | 在预算委员会中,我们对 2011 年预算草案的 1600 多项修正案进行了表决……并在全体会议上进行了表决。 |
| CohereLabs/cohere-transcribe-03-2026 | 主席先生,在预算委员会中,我们对 2011 年预算草案的 1 项以上修正案进行了表决……并在全体会议上进行了表决。 |
| nvidia/canary-qwen-2.5b | 主席先生,在预算委员会中,我们对 2011 年预算草案的 1 项以上修正案进行了表决……并在全体会议上进行了表决。 |
| ibm-granite/granite-speech-4.1-2b | 主席先生,在预算委员会中,我们对 2011 年预算草案的 1600 多项修正案进行了表决……并在全体会议上进行了表决。 |
| microsoft/Phi-4-multimodal-instruct | 主席先生,在预算委员会中,我们对 2011 年预算草案的 1 项以上修正案进行了表决……并在全体会议上进行了表决。 |
| nvidia/parakeet-tdt-0.6b-v2 | 主席先生,在预算委员会中,我们对 2011 年预算草案的 1 项以上修正案进行了表决……并在新教徒中进行了表决。 |
| bosonai/higgs-audio-v3-8b-stt-v2 | 主席先生,在预算委员会中,我们对 2011 年预算草案的 1600 多项修正案进行了表决……并在全体会议上进行了表决。 |
| Qwen/Qwen3-ASR-0.6B-hf | 主席先生,在预算委员会中,我们对 2011 年预算草案的 1600 多项修正案进行了表决……并在全体会议上进行了表决。 |
| mistralai/Voxtral-Mini-3B-2507 | 主席先生,在预算委员会中,我们对 2011 年预算草案的 1600 多项修正案进行了表决……并在全体会议上进行了表决。 |
| moonshotai/Kimi-Audio-7B-Instruct | 主席先生,在预算委员会中,我们对 2011 年预算草案的 1600 多项修正案进行了表决……并在全体会议上进行了表决。 |
| openai/whisper-large-v3 | 主席先生,在预算委员会中,我们对2011年预算草案的1600多项修正案进行了表决……并在全体会议上进行了表决 |
| moonshine-ai/moonshine-streaming-medium | 主席先生,在预算委员会中,我们对2011年预算草案的一千六百多项修正案进行了表决……并在全体会议上进行了表决 |
在公开基准测试上,恢复率最高,而在保留集或新采集的音频(如下方的 ep-fresh 和 libri-fresh)上恢复率较低。在 LibriSpeech 上,一些表现最强的基准模型在大约 30–40% 的样本中重现了被遮蔽的数字,尽管该数字本身已被移除。对于几个模型而言,这种效应在新采集的数据上有所减弱,这表明与基准测试相关的周边音频——而不仅仅是文本自动补全——帮助模型恢复了参考内容。

拼写切换
我们的拼写切换探针测试的是:模型是否会重现基准测试参考转写中使用的确切拼写,尽管音频中该拼写并不清晰。拼写变体是指语义和发音完全相同、但拼写方式不同的单词(如 1 与 one、Mr. 与 mister、John 与 Jon、Honor 与 Honour 等)。理论上,模型应当始终偏好某一种拼写,或者以大致随机的比例在两种拼写之间切换。如果模型系统性地切换以匹配每个基准测试参考转写中的拼写,那就表明模型捕捉到了测试所期望的拼写方式。
在 LibriSpeech 中,我们测试了一种数据集内部的切换,涉及一种较旧的空格约定:一些参考转写使用“any one”,而另一些则使用“anyone”。我们测量给定变体的最低准确率,称之为“切换率”。如果模型只使用一种变体,其切换率就是 0%;如果模型随机选择,预期切换率约为 50%;如果模型在每个测试样本中都知道该使用哪种变体,则切换率可达 100%。

我们的第二个探针测试的是数据集之间的切换,即每个基准在其整个测试语料库中一致地使用不同的拼写约定。例如,VoxPopuli 使用缩写“Mr.”,而 LibriSpeech 则拼出完整的“Mister”。
多个模型超过了 50% 的随机选择基线,其中一些模型的切换准确率接近 90%。这表明,模型能够识别音频样本来自哪个数据集,并选择该基准所期望的拼写约定,尽管这两种形式听起来完全相同。

定位切换行为
为了测试这些行为是否能泛化到公开基准之外,我们还从相同来源领域收集了新数据,但这些数据是在模型训练截止日期之后产生的:包括 VoxPopuli 对应的近期欧洲议会录音,以及 LibriSpeech 对应的新活跃 LibriVox 朗读者录音。然而,当面对来自同一领域的新采集数据时,许多模型不再匹配参考转写,而是回归到更忠实于音频的转写。
其他干预措施也指向同样的结论。音频中存在但在参考转写中被省略的短语,在要求模型翻译音频或将其注意力限制在相关帧时,可能会重新出现。裁剪掉周围的基准上下文,或附加普通的对话音频,也能恢复忠实的转写。而附加 VoxPopuli 音频则可能产生相反的效果,使原本忠实的合成或挖掘样本更有可能匹配基准参考。

综合来看,这些结果表明,模型能够忠实地转写字面上的口语词汇,但会利用周围的声学上下文来决定是遵循音频内容,还是遵循特定基准的转写策略。
结论
我们的研究结果表明,在两个主要的开源数据集上,部分模型能够检测到与数据集相关的声学线索,并据此调整其转写行为。具体而言,模型可能会复现音频中不存在但参考转写文本中存在的词语,以较高概率恢复被静音的数字,或利用周围的声学上下文来选择特定基准所期望的书面变体。
对于正在挑选模型的用户而言,这些发现凸显了使用完全留出的评估集(如 RW-Voice-EQ Bench 和 Open ASR Leaderboard 所做的那样)的重要性,以及不能仅关注单一公开基准上的词错误率。为此,Open ASR Leaderboard 新增了一个“基准拟合”标签页,其中包含针对所有模型的上述两项分析:量化(1)来自 VoxPopuli 的参考错误率,以及(2)所有公开数据集上的拼写切换情况。相关脚本以及未归一化的模型输出已在 GitHub 上开源。
我们的研究结果还表明,基准开发者应避免使用简单的独立同分布测试集划分,而应采用基于时间、说话者或其他元数据的划分方式。训练数据和模型选择流程的更高透明度,也将有助于研究人员理解这些行为是如何产生的。
公开基准仍然具有价值:它们透明、可重复、易于运行,并且被研究社区广泛理解。但只有在能够区分真正的转写改进与无法泛化到新音频的基准特定收益时,它们才最为有用。
如需了解更多信息,我们建议您阅读我们的完整报告。
Public voice AI benchmarks increasingly suggest that models are performing at human levels. Yet those scores don't always reflect how models work in the real-world. Since public benchmarks are open and widely used, models can also become optimized for the tests themselves. Their scores may improve because they have learned benchmark-specific patterns and not because they have become better at the underlying task. One reason is that traditional benchmarks overlook many of the conditions and qualities that make voice systems reliable, natural, contextually appropriate, and effective in practice. That's why we recently introduced held-out sets in Real World VoiceEQ, the Open-ASR Leaderboard, and the Far-field ASR Leaderboard: to measure more of what matters in real-world use.
However, broader measurement alone does not solve the problem. This phenomenon, sometimes called benchmark optimization or "benchmaxxing," is often discussed around machine learning, however, it has been difficult to measure in speech recognition.
Our latest research introduces three tests to help quantify it. We evaluated 11 widely used open-source ASR models and found that several of the highest-scoring systems reproduced benchmark transcripts from the VoxPopuli English and LibriSpeech (clean, other) datasets – even when the audio contradicted them, relevant words had been silenced, or the audio equally supported two different written forms.
In some cases, models appeared to rely not only on what was said, but also on subtle acoustic cues that indicated which benchmark they were being tested on. As a result, their scores overstated how well they could transcribe speech more generally.
Reference disagreement (VoxPopuli case study)
VoxPopuli is known to contain a high number of transcription errors (which is why Artificial Analysis released a cleaned version). Our consensus disagreement probe tests what happens when leading ASR models encounter these errors: Do they accurately transcribe what the audio says, or reproduce the benchmark's incorrect reference transcript?
To test this at scale, we use an ensemble of independent models selected for their low phoneme error rate (PER). PER measures how closely a written transcription matches the sounds in the audio, making it a useful proxy for how faithfully a model transcribes what it hears. The ensemble results can be used to flag cases in which the models unanimously disagree with the benchmark's reference transcript. We then compare a sample of those flagged cases against human annotations to validate the corrected transcripts.
For example, one VoxPopuli clip audibly includes the phrase "Thank you, Mr. President," but the reference transcript omits "Thank you." Six of the 11 models we tested reproduced the benchmark's erroneous transcript—giving the "expected" answer even though it contradicted the audio. On the real clip, the formatting follows the same pattern: models that omit "Thank you" also reproduce the benchmark's punctuation style, writing "Mr" without a period, while models that include the audible phrase tend to write "Mr." with the period.
When we present the same content in newly collected voices from EU parliamentary recordings or generic voices, this behavior often weakens or disappears. In the below samples, all but one model flips back to transcribing the audio-faithful transcript for a clone of a new parliamentary recording. This suggests that the models are responding to acoustic cues that help them identify the benchmark membership and thus produce the expected transcript even if it contradicts the audio.
The reference transcript for this clip reads "Mr President, I have another complaint about this procedure, which is that it is not secret." The audio in all three clips below actually says the same thing, preceded by an audible "Thank you,"—the clones are text-to-speech renditions of that true sentence, so the courtesy is audible in all three. Green highlighting and ✅ mark a transcript that includes the audible "Thank you"; red highlighting and ❌ mark a transcript that reproduces the benchmark's erroneous omission. All transcripts are raw model output, prior to any normalization—casing and punctuation are preserved exactly as generated, including lowercase output from some models.
Original VoxPopuli recording
Voice clone of the same speaker
Clone of a parliament speaker recorded after every model's training cutoff
| Model | Real clip | Same-speaker clone | ep-fresh clone |
|---|---|---|---|
| CohereLabs/cohere-transcribe-03-2026 | ❌ Mr President… | ❌ Mr President… | ✅ Thank you, Mr President… |
| nvidia/canary-qwen-2.5b | ❌ Mr President… | ❌ Mr President… | ✅ Thank you Mr. President… |
| ibm-granite/granite-speech-4.1-2b | ❌ mr president… | ❌ mr president… | ✅ thank you mr president… |
| microsoft/Phi-4-multimodal-instruct | ❌ Mr President… | ❌ Mr President… | ❌ Mr President… |
| nvidia/parakeet-tdt-0.6b-v2 | ❌ Mr President… | ✅ Thank you, Mr President… | ✅ Thank you, Mr. President… |
| bosonai/higgs-audio-v3-8b-stt-v2 | ❌ mr president… | ❌ mr president… | ✅ thank you mr president… |
| Qwen/Qwen3-ASR-0.6B-hf | ✅ Thank you, Mr. President… | ✅ Thank you, Mister President… | ✅ Thank you, Mister President… |
| mistralai/Voxtral-Mini-3B-2507 | ✅ Thank you, Mr. President… | ✅ Thank you, Mr. President… | ✅ Thank you, Mr. President… |
| moonshotai/Kimi-Audio-7B-Instruct | ✅ Thank you, mr. President… | ✅ Thank you, Mr. President… | ✅ Thank you, mr. President… |
| openai/whisper-large-v3 | ✅ Thank you, Mr. President… | ✅ Thank you, Mr. President… | ✅ Thank you, Mr. President… |
| moonshine-ai/moonshine-streaming-medium | ✅ thank you mr president… | ✅ thank you mr president… | ✅ thank you mr president… |
| Drops the courtesy (❌) out of 11 | 6 | 5 | 1 |
Parakeet is the only model that flips between reproducing the benchmark on the real clip and getting it right on the same-speaker clone. Phi-4 is the only model still dropping the courtesy on the ep-fresh clone. When we instead resynthesize the sentence in a generic TTS voice unconnected to any parliamentary recording, all eleven models restore the courtesy.
The results suggest that this problem is both widespread and meaningful. Our methodology flagged potential reference errors in 40% of the VoxPopuli test clips we analyzed, affecting roughly 3% of all reference words.
Models exhibiting benchmark-optimized behavior reproduced erroneous reference transcripts 18–30% of the time. The scatterplot below compares VoxPopuli word error rate (WER) on the x-axis with the rate at which each model reproduces the benchmark's incorrect reference instead of the consensus correction. The models with the lowest WER—and therefore the strongest reported benchmark performance—are also the most likely to reproduce these errors.

Masked Entity Retrieval
To build on the consensus disagreement probe, we deliberately silence numbers in the audio samples of test datasets and ask the models to transcribe what it hears. The number is literally absent from the audio, so models should not output any number, much less the exact number in the text.
Some of these numbers are semi-predictable (although still unlikely for a model to predict), yet others are quite surprising. The following clip combines both probes, showing both how models recreate reference transcript errors including an incorrect number and one model even autocompletes a relatively random year (2011) despite it being silenced. In each model's row below:
- green highlighting with strikethrough marks reference-transcript words the model correctly did not reproduce (audio-faithful);
- green highlighting with underline marks a correct, audio-faithful insertion in place of the reference's erroneous wording;
- red highlighting (plain text) reproduces the reference transcript's erroneous, audio-unsupported content: keeping "Mr President", writing "more than 1 amendments" where the audio says "one thousand six hundred", supplying the silenced year "2011", or ending on "plenary".
2011 draft budget (masked numbers)
| Reference | Mr President, in the Committee on Budgets, we voted on more than 1 amendments to the 2011 draft budget … voted in the plenary. |
|---|---|
| What the audio says | In the Committee on Budgets, we voted on more than one thousand six hundred amendments to the ⟨silenced⟩ draft budget … voted in the … |
| CohereLabs/cohere-transcribe-03-2026 | Mr President, in the Committee on Budgets we voted on more than 1 amendments to the 2011 draft budget … voted in the plenary. |
| nvidia/canary-qwen-2.5b | Mr President, in the Committee on Budgets we voted on more than one amendments to the |
| ibm-granite/granite-speech-4.1-2b | |
| microsoft/Phi-4-multimodal-instruct | |
| nvidia/parakeet-tdt-0.6b-v2 | |
| bosonai/higgs-audio-v3-8b-stt-v2 | |
| Qwen/Qwen3-ASR-0.6B-hf | |
| mistralai/Voxtral-Mini-3B-2507 | |
| moonshotai/Kimi-Audio-7B-Instruct | |
| openai/whisper-large-v3 | |
| moonshine-ai/moonshine-streaming-medium |
Recovery rates were highest on the public benchmarks and lower on held-out or newly collected audio (ep-fresh and libri-fresh below). On LibriSpeech, some of the strongest benchmark-performing models reproduced masked numbers in roughly 30–40% of examples, even though the number itself had been removed. The effect weakened on freshly collected data for several models, suggesting that the surrounding benchmark-associated audio—not only textual autocomplete—helped the models recover the reference.

Orthographic Switching
Our orthographic switching probe tests whether models reproduce the exact spelling used in a benchmark's reference transcript despite it not being clear in the audio. Orthographic variants are words that are semantically and phonetically identical but can be spelled different ways (1 vs one, Mr. vs mister, John vs Jon, Honor vs Honour, etc). In theory, models should consistently prefer one spelling over another, or alternate between them at roughly random rates. If models systematically switch to match what is in each benchmark's reference transcript, that suggests the models are picking up on which spelling the test expects.
Transcription: "I URGED ON THE BOYS THAT WHATEVER HAPPENED WE SHOULD NOT SHOOT ANY ONE" — models using "any one": 6/11, models using "anyone": 5/11
Transcription: "CAMOUFLAGE WAS NOT A WORD THE CAPTAIN OR ANYONE ELSE OF HIS TIME YET UNDERSTOOD" — models using "any one": 2/11, models using "anyone": 9/11
Within LibriSpeech, we test one intra-dataset switch involving an older spacing convention: some reference transcripts use "any one", while others use "anyone." We measure the minimum accuracy for a given variant, which we call "switch rate". If a model only uses one variant it would have a 0% switch rate; a model which picks randomly would be expected to have a 50% switch rate. A model which knows which variant to use in every test sample would earn a 100% switch rate.

Our second probe tests an inter-dataset switch, in which each benchmark uses a different spelling convention consistently across its test corpus. For example, VoxPopuli uses the abbreviation "Mr.," while LibriSpeech spells out "Mister."
Multiple models exceed the 50% random-choice baseline, with some reaching roughly 90% switch accuracy. This suggests that the models can identify which dataset an audio sample comes from and select the spelling convention that benchmark expects, even though both forms sound identical.

Localizing the switches
To test whether these behaviors generalize beyond the public benchmarks, we also collected fresh data from the same source domains but after the models' training cutoffs: recent European Parliament recordings for VoxPopuli and recordings from newly active LibriVox narrators for LibriSpeech. However, when presented with recently collected data from the same domain, many models stop matching the reference transcript and revert to more audio faithful transcriptions.
Other interventions point to the same conclusion. Phrases which are present in the audio but are omitted in the reference transcript can reappear when a model is asked to translate the audio or when its attention is restricted to the relevant frames. Trimming away surrounding benchmark context, or appending ordinary conversational audio, can also restore the faithful transcript. Appending VoxPopuli audio can have the opposite effect, making otherwise faithful synthetic or mined samples more likely to match the benchmark reference.

Together, these results suggest that models are able to faithfully transcribe the literal spoken words, but are using surrounding acoustic context to decide whether to follow the audio or a benchmark-specific transcription policy.
Conclusion
Our findings suggest that, on two major open-source datasets, some models detect dataset-associated acoustic cues and adjust their transcription behavior accordingly. Specifically, models may reproduce words that are absent from the audio but present in the reference transcript, recover silenced numbers at elevated rates, or use surrounding acoustic context to select the written variant expected by a particular benchmark.
For people selecting models, these findings underscore the importance of using fully held-out evaluation sets, as RW-Voice-EQ Bench and the Open ASR Leaderboard do, and of looking beyond word error rate on a single public benchmark. To this end, a "Benchmark fitting" tab has been added to the Open ASR Leaderboard, which includes two of the above analyses across all models: quantifying (1) reference error rates from VoxPopuli and (2) orthographic switching across all public datasets. The relevant scripts are open-sourced on GitHub as well as the un-normalized model outputs.
Our findings also suggest that benchmark developers should avoid simple independent and identically distributed test splits in favor of temporal, speaker, or other metadata-based separation. Greater transparency around training data and model-selection procedures would also help researchers understand how these behaviors arise.
Public benchmarks remain valuable: they are transparent, repeatable, easy to run, and well understood by the research community. But they are most useful when we can distinguish genuine transcription improvements from benchmark-specific gains that do not generalize to new audio.
For more information, we encourage you to read our full report.