Dataset Viewer
Auto-converted to Parquet
prompt
stringlengths
118
420
run_1_overfit
float64
17.4
196
llama_prompt_epoch5
float64
7.56
78.2
llama_prompt_epoch2
float64
8.49
56.7
chunk_0-5s_epoch5
float64
1.19
1.86
chunk_0-25s_epoch5
float64
1.2
1.9
chunk_0-5s_epoch2
float64
1.18
1.73
chunk_0-25s_epoch2
float64
1.2
1.92
llama-3.2-1b_original
float64
6.33
9.42
llama-3.1-8b_original
float64
5.47
7.66
llama-3.1-8b_score0-10
stringlengths
256
439
gpt-5_score0-10
stringlengths
232
371
gemini-2.5-flash_score0-10
stringlengths
235
378
<|TIME_TOKEN|>Did you<|SPEAKER_SEP|> <|TIME_TOKEN|>eat lunch?<|SPEAKER_SEP|>Yeah, <|TIME_TOKEN|><|SPEAKER_SEP|>just now.
20.803404
9.441545
10.894551
1.191159
1.201747
1.182045
1.198785
9.152042
7.392473
{ "fluency": 8, "coherence": 9, "timing_realism": 7, "engagement": 6, "overall": 7, "comments": "Coherent and grammatically correct, but somewhat abrupt turns. Timing is a bit off with the short responses, but still feels somewhat realistic." }
{ "fluency": 9, "coherence": 10, "timing_realism": 10, "engagement": 8, "overall": 9, "comments": "Natural, concise exchange with clear question-answer alignment. Realistic overlap and silences across time tokens. Minor brevity/fragmentation prevents a perfect score." }
```json { "fluency": 10, "coherence": 10, "timing_realism": 10, "engagement": 9, "overall": 9, "comments": "A very brief but perfectly natural and well-timed exchange. The slight overlap is realistic, and the coherence is excellent." } ```
<|TIME_TOKEN|>Did you<|SPEAKER_SEP|> <|TIME_TOKEN|>eat lunch?<|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|>Yeah, <|TIME_TOKEN|><|SPEAKER_SEP|>just now.
18.026825
7.560242
8.492839
1.276782
1.31087
1.232844
1.296968
8.749963
7.150039
{ "fluency": 8, "coherence": 9, "timing_realism": 7, "engagement": 6, "overall": 7, "comments": "Simple and coherent dialogue. Good fluency, but the timing could be more realistic, with slightly awkward overlaps. Engagement is somewhat low due to the short and straightforward nature of the conversation." }
{ "fluency": 10, "coherence": 10, "timing_realism": 8, "engagement": 6, "overall": 8, "comments": "Natural, grammatically correct Q&A with clear relevance. Timing is plausible with clean turn-taking, though simple and brief, which limits engagement." }
```json { "fluency": 10, "coherence": 10, "timing_realism": 9, "engagement": 9, "overall": 9, "comments": "A very short but perfectly fluent and coherent exchange. The timing for the question and answer is realistic." } ```
<|TIME_TOKEN|>Did you<|SPEAKER_SEP|> <|TIME_TOKEN|>eat lunch?<|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|>Yeah, <|TIME_TOKEN|><|SPEAKER_SEP|>just now.
27.408072
12.928969
11.527799
1.346788
1.388518
1.306672
1.371211
8.521586
7.015492
{ "fluency": 8, "coherence": 7, "timing_realism": 6, "engagement": 6, "overall": 7, "comments": "The dialogue is fairly fluent, but there's a brief pause between the first two turns. Coherence is somewhat compromised by the abrupt shift in topic. Timing and engagement are somewhat lacking due to the short and somewhat disjointed nature of the conversation." }
{ "fluency": 10, "coherence": 10, "timing_realism": 9, "engagement": 5, "overall": 8, "comments": "Clear, fluent Q&A with a natural brief pause before the reply. Well-aligned turns but very short, limiting engagement." }
```json { "fluency": 10, "coherence": 10, "timing_realism": 9, "engagement": 8, "overall": 9, "comments": "A very short but perfectly fluent and coherent exchange. The timing is natural, with a brief pause before the answer." } ```
<|TIME_TOKEN|>Did you<|SPEAKER_SEP|> <|TIME_TOKEN|>eat lunch?<|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|>Yeah, <|TIME_TOKEN|><|SPEAKER_SEP|>just now.
35.322468
19.547939
13.985072
1.424435
1.461759
1.374409
1.44072
8.257158
6.838672
{ "fluency": 6, "coherence": 7, "timing_realism": 5, "engagement": 6, "overall": 6, "comments": "Dialogue has some fluency issues due to incomplete sentences. Coherence is mostly maintained, but the timing feels a bit forced with multiple consecutive empty turns. Engagement is moderate, but the conversation feels somewhat one-sided." }
{ "fluency": 10, "coherence": 10, "timing_realism": 9, "engagement": 6, "overall": 8, "comments": "Natural and grammatically correct exchange with a coherent reply. Timing feels realistic with a brief pause before the response. Very short interaction limits engagement." }
```json { "fluency": 10, "coherence": 10, "timing_realism": 10, "engagement": 9, "overall": 9, "comments": "A very brief but perfectly natural and coherent exchange. The timing, including the short pause before the answer, feels very realistic." } ```
<|TIME_TOKEN|>Did you<|SPEAKER_SEP|> <|TIME_TOKEN|>eat lunch?<|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|>Yeah, <|TIME_TOKEN|><|SPEAKER_SEP|>just now.
47.52951
24.914587
14.934292
1.493432
1.523001
1.433244
1.507778
7.971219
6.631482
{ "fluency": 6, "coherence": 8, "timing_realism": 4, "engagement": 6, "overall": 6, "comments": "The dialogue is generally coherent, but fluency is somewhat affected by the incomplete and repetitive turns. Timing and realism suffer from the excessive use of empty turns, making the conversation feel somewhat stilted." }
{ "fluency": 10, "coherence": 10, "timing_realism": 9, "engagement": 4, "overall": 8, "comments": "Natural, grammatically correct Q&A with clear alignment; realistic pause before the reply. Very brief exchange limits engagement." }
```json { "fluency": 10, "coherence": 10, "timing_realism": 8, "engagement": 8, "overall": 9, "comments": "The dialogue is perfectly fluent and coherent. The timing includes a noticeable pause before the response, which is realistic, though perhaps a bit long for a simple question." } ```
<|TIME_TOKEN|>Did you<|SPEAKER_SEP|> <|TIME_TOKEN|>eat lunch?<|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|>Yeah, <|TIME_TOKEN|><|SPEAKER_SEP|>just now.
55.948055
28.567921
14.901108
1.552421
1.58613
1.481601
1.576291
7.687777
6.427553
{ "fluency": 6, "coherence": 8, "timing_realism": 4, "engagement": 6, "overall": 6, "comments": "The dialogue is somewhat coherent, but the fluency is affected by the repetitive and empty turns from speaker A. The timing and realism are also lacking due to the excessive use of <|TIME_TOKEN|> and <|SPEAKER_SEP|> without any meaningful content." }
{ "fluency": 9, "coherence": 10, "timing_realism": 9, "engagement": 6, "overall": 8, "comments": "Clear, grammatical turns with a coherent Q&A. The pause before the reply feels realistic, and the staggered response is natural. Very brief exchange limits engagement." }
```json { "fluency": 10, "coherence": 10, "timing_realism": 7, "engagement": 7, "overall": 8, "comments": "The dialogue is fluent and coherent. However, there's a significant pause from Speaker B before responding, which impacts the realism of the timing and the immediate engagement." } ```
<|TIME_TOKEN|>Did you<|SPEAKER_SEP|> <|TIME_TOKEN|>eat lunch?<|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|>Yeah, <|TIME_TOKEN|><|SPEAKER_SEP|>just now.
31.403473
30.726828
11.064485
1.787825
1.889059
1.671838
1.9224
6.504412
5.536175
{ "fluency": 4, "coherence": 2, "timing_realism": 3, "engagement": 2, "overall": 2, "comments": "The dialogue is marred by excessive use of empty turns (separated by <|SPEAKER_SEP|> without any content), disrupting the flow and coherence. The engagement is low due to the lack of meaningful responses. Timing and realism are also compromised by the numerous empty turns." }
{ "fluency": 10, "coherence": 10, "timing_realism": 6, "engagement": 3, "overall": 6, "comments": "Natural and coherent exchange with a direct answer to the question. However, an extended stretch of silence and minimal back-and-forth reduce realism and engagement." }
```json { "fluency": 10, "coherence": 10, "timing_realism": 2, "engagement": 5, "overall": 6, "comments": "The dialogue is coherent and fluent, but the extremely long pause between the question and the answer makes the timing highly unrealistic and diminishes engagement." } ```
<|TIME_TOKEN|>Did you<|SPEAKER_SEP|> <|TIME_TOKEN|>eat lunch?<|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|>Yeah, <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|>just now.
41.870129
47.355377
19.262724
1.859187
1.900185
1.729072
1.913841
6.510108
5.567114
{ "fluency": 6, "coherence": 4, "timing_realism": 2, "engagement": 3, "overall": 4, "comments": "The dialogue lacks fluency due to the excessive use of empty tokens and awkward pauses. Coherence is also compromised by the abrupt and unrelated responses. Timing and realism are severely affected by the unnatural and prolonged silences. Engagement is low due to the lack of meaningful interaction and response from speaker B." }
{ "fluency": 9, "coherence": 9, "timing_realism": 6, "engagement": 3, "overall": 6, "comments": "Clear, natural phrasing and coherent Q&A. However, long stretches of silence and minimal back-and-forth reduce timing realism and engagement." }
```json { "fluency": 10, "coherence": 10, "timing_realism": 4, "engagement": 5, "overall": 6, "comments": "The utterances are perfectly fluent and coherent. However, the extensive empty time tokens between the question and answer create an unrealistic and excessively long pause, significantly impacting timing and engagement." } ```
<|TIME_TOKEN|>Did you<|SPEAKER_SEP|> <|TIME_TOKEN|>eat lunch?<|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|> <|TIME_TOKEN|><|SPEAKER_SEP|>
32.049149
35.137215
9.763943
1.724045
1.835408
1.608769
1.864116
6.329377
5.474041
{ "fluency": 6, "coherence": 5, "timing_realism": 4, "engagement": 4, "overall": 5, "comments": "The dialogue lacks coherence and engagement. The repeated speaker separation and time tokens create an unnatural and awkward flow. The conversation seems to be stuck in an incomplete state." }
{ "fluency": 9, "coherence": 2, "timing_realism": 3, "engagement": 1, "overall": 2, "comments": "A clear, fluent question split over two time steps, but the second speaker never responds and the dialogue stalls with many silent intervals, resulting in low coherence, realism, and engagement." }
{{ "fluency": 8, "coherence": 2, "timing_realism": 1, "engagement": 1, "overall": 2, "comments": "The dialogue consists of one speaker asking a question, followed by a lengthy and unrealistic silence from the second speaker. This prevents any assessment of coherence, engagement, or realistic turn-taking beyond the initial utterance." }}
<|TIME_TOKEN|>Did you<|SPEAKER_SEP|> <|TIME_TOKEN|>eat lunch?<|SPEAKER_SEP|>Machine learning <|TIME_TOKEN|><|SPEAKER_SEP|>is intersting.
195.917435
78.152916
56.736393
1.297761
1.309508
1.264881
1.293246
9.418255
7.66288
{ "fluency": 6, "coherence": 4, "timing_realism": 5, "engagement": 3, "overall": 4, "comments": "The dialogue starts off with a coherent question, but speaker B's response is unrelated to the topic of eating lunch. The timing is somewhat unnatural, with a large gap between the question and the response." }
{ "fluency": 6, "coherence": 2, "timing_realism": 5, "engagement": 2, "overall": 3, "comments": "Speaker A asks about lunch while Speaker B talks about machine learning, so they talk past each other. Fluency is moderate with one spelling error ('intersting'). Timing markers show some pauses/overlap but the interaction lacks responsiveness and engagement." }
```json { "fluency": 10, "coherence": 0, "timing_realism": 3, "engagement": 0, "overall": 1, "comments": "Speaker B's response is completely irrelevant to Speaker A's question, leading to a total breakdown in coherence and engagement. While the individual phrases are fluent and timing shows an overlap, the content mismatch makes the interaction unrealistic." } ```
<|TIME_TOKEN|>Did you<|SPEAKER_SEP|> <|TIME_TOKEN|>eat lunch?<|SPEAKER_SEP|>I like <|TIME_TOKEN|><|SPEAKER_SEP|>pizza.
31.212111
16.993347
15.922923
1.199354
1.205628
1.188581
1.213451
9.150151
7.485902
{ "fluency": 8, "coherence": 7, "timing_realism": 6, "engagement": 7, "overall": 7, "comments": "The dialogue is mostly fluent, but the speaker B's response seems a bit abrupt and lacks context. The timing and realism could be improved with more natural overlaps or pauses." }
{ "fluency": 9, "coherence": 6, "timing_realism": 7, "engagement": 4, "overall": 6, "comments": "Grammatically natural utterances. B’s response is related but doesn’t directly answer the question, reducing coherence. Timing includes a mild overlap that feels plausible, though the split phrasing is a bit choppy. Very short exchange limits engagement." }
```json { "fluency": 9, "coherence": 1, "timing_realism": 7, "engagement": 1, "overall": 2, "comments": "Speaker B's response is completely irrelevant to Speaker A's question, making the dialogue incoherent and lacking in engagement, despite the individual phrases being fluent." } ```
<|TIME_TOKEN|>Did you<|SPEAKER_SEP|> <|TIME_TOKEN|>eat lunch?<|SPEAKER_SEP|>It's a <|TIME_TOKEN|><|SPEAKER_SEP|>sunny day.
17.445253
12.192446
12.937063
1.225546
1.22762
1.2032
1.228357
9.123083
7.534726
{ "fluency": 8, "coherence": 6, "timing_realism": 7, "engagement": 4, "overall": 6, "comments": "Fluency is good, but the dialogue feels a bit abrupt and lacks contextual relevance. The mention of'sunny day' seems unrelated to the initial question about eating lunch." }
{ "fluency": 9, "coherence": 2, "timing_realism": 7, "engagement": 3, "overall": 4, "comments": "Utterances are grammatically correct and timing shows plausible overlap, but the speakers talk past each other. The weather remark does not respond to the lunch question, resulting in low coherence and engagement." }
{{ "fluency": 9, "coherence": 1, "timing_realism": 6, "engagement": 1, "overall": 1, "comments": "The dialogue completely lacks coherence and engagement as Speaker B's response is entirely irrelevant to Speaker A's question. While individual utterances are fluent, the interaction fails." }}
<|TIME_TOKEN|>Did you<|SPEAKER_SEP|> <|TIME_TOKEN|>eat lunch?<|SPEAKER_SEP|>Yes, I've <|TIME_TOKEN|><|SPEAKER_SEP|>just have pasta.
25.771673
15.981815
16.873919
1.226028
1.25014
1.222534
1.248949
9.210662
7.62244
{ "fluency": 8, "coherence": 9, "timing_realism": 7, "engagement": 7, "overall": 8, "comments": "Dialogue is coherent and grammatically correct, but the timing feels a bit rushed. Speakers' turns are short and lack some natural overlap or pause." }
{ "fluency": 5, "coherence": 8, "timing_realism": 9, "engagement": 6, "overall": 7, "comments": "Natural overlap and clear Q/A alignment. Fluency suffers due to B's grammatical error ('Yes, I've ... just have pasta'), and the exchange is very brief, limiting engagement." }
```json { "fluency": 6, "coherence": 10, "timing_realism": 9, "engagement": 7, "overall": 7, "comments": "The dialogue is coherent and has realistic timing, with a natural overlap. However, Speaker B's response contains a grammatical error ('I've just have pasta' should be 'I've just had pasta'), which impacts fluency." } ```
<|TIME_TOKEN|>Did you<|SPEAKER_SEP|> <|TIME_TOKEN|>eat lunch?<|SPEAKER_SEP|>How's the <|TIME_TOKEN|><|SPEAKER_SEP|>weather today?
34.523445
14.688357
15.434488
1.203414
1.215006
1.204507
1.213478
9.13752
7.470387
{ "fluency": 8, "coherence": 7, "timing_realism": 6, "engagement": 6, "overall": 7, "comments": "The dialogue is somewhat natural, but the turns are a bit abrupt and lack fluidity. The coherence is decent, but the topics are somewhat disconnected. Timing and realism are not well-represented, with a lack of realistic overlaps or silences." }
{ "fluency": 9, "coherence": 3, "timing_realism": 8, "engagement": 3, "overall": 5, "comments": "Both sides produce grammatically correct, natural questions. However, they talk past each other with no direct responses, hurting coherence and engagement. The timing shows realistic overlap and turn distribution across tokens." }
```json { "fluency": 9, "coherence": 1, "timing_realism": 5, "engagement": 2, "overall": 2, "comments": "The dialogue lacks coherence as Speaker B completely disregards Speaker A's question and introduces an entirely new topic, making it feel like two separate monologues rather than a conversation. This severely impacts engagement." } ```
<|TIME_TOKEN|>Did you<|SPEAKER_SEP|> <|TIME_TOKEN|>eat lunch?<|SPEAKER_SEP|>No, I don’t <|TIME_TOKEN|><|SPEAKER_SEP|>actually eat.
48.789009
32.721718
27.633865
1.267966
1.278192
1.245497
1.271359
9.244843
7.627542
{ "fluency": 8, "coherence": 7, "timing_realism": 6, "engagement": 6, "overall": 7, "comments": "Dialogue is somewhat coherent, but the fluency could be improved with more natural phrasing. Timing and engagement are a bit off, with a slightly awkward turn-taking rhythm." }
{ "fluency": 10, "coherence": 10, "timing_realism": 9, "engagement": 7, "overall": 9, "comments": "Natural, grammatically sound exchange with a coherent reply and believable slight overlap; brief length limits engagement." }
```json { "fluency": 10, "coherence": 10, "timing_realism": 8, "engagement": 7, "overall": 8, "comments": "A short but perfectly fluent and coherent exchange. Speaker B's response overlaps naturally with A's question, though A's speech is broken into slightly small segments." } ```
README.md exists but content is empty.
Downloads last month
97