id stringlengths 10 10 | clip_id stringclasses 6
values | audio audioduration (s) 6.36 11.3 | task stringlengths 18 40 | family stringclasses 6
values | question stringlengths 36 91 | context stringlengths 201 2.94k | options listlengths 3 11 | answer stringlengths 2 23 | current_node stringclasses 3
values | duration_seconds float64 6.36 11.3 | duration_bucket stringclasses 2
values |
|---|---|---|---|---|---|---|---|---|---|---|---|
sample-001 | example-01 | conversation.admission_interest | Business and conversation | What programme interest does the caller express at this point? | The agent is checking whether the caller is interested in the programme. The agent is the AI assistant; the caller is the human/user. Use only this clip and the supplied context through the clip's end. | [
{
"id": "interested",
"description": "Interested: Clear affirmative interest in the admission/program under discussion."
},
{
"id": "not_interested",
"description": "Not interested: Explicit refusal or lack of interest in that offering."
},
{
"id": "conditional_interest",
"descriptio... | interested | 6.36 | 05-10s | ||
sample-002 | example-01 | business.cf_ai_lead_qualification_status | Business and conversation | What lead qualification status is supported under the supplied business rubric? | The agent is checking whether the caller is interested in the programme. The agent is the AI assistant; the caller is the human/user. Use only this clip and the supplied context through the clip's end.
Business rubric:
Determine whether the lead is qualified based on the conversation.
Return "Qualified" if the parent... | [
{
"id": "Qualified",
"description": "Qualified: Select 'Qualified' exactly under the original field rubric and allowed-output definitions in context. Apply it only to the observed prefix through the clip cutoff, not the eventual whole-call disposition. Source option definition: user shows interest OR asks q... | Qualified | 6.36 | 05-10s | ||
sample-003 | example-02 | conversation.availability | Business and conversation | Is the caller available to talk now? | Current node: opening. The agent is checking whether the caller has time for the conversation. The agent is the AI assistant; the caller is the human/user. Use only this clip and the supplied context through the clip's end. | [
{
"id": "available_now",
"description": "Available now: The person agrees to talk or clearly participates now."
},
{
"id": "busy_or_unavailable",
"description": "Busy or unavailable: They cannot talk now; this alone does not establish outbound callback consent."
},
{
"id": "asks_to_wait"... | available_now | 7.02 | 05-10s | ||
sample-004 | example-02 | business.cf_ai_callback_request | Business and conversation | Has the caller requested or accepted a callback or human follow-up? | Current node: opening. The agent is checking whether the caller has time for the conversation. The agent is the AI assistant; the caller is the human/user. Use only this clip and the supplied context through the clip's end.
Business rubric:
Determine whether the caller requested a callback or asked to speak with a hum... | [
{
"id": "Yes",
"description": "Yes: Select 'Yes' exactly under the original field rubric and allowed-output definitions in context. Apply it only to the observed prefix through the clip cutoff, not the eventual whole-call disposition. Source option definition: \"call me later\" OR \"call tomorrow\" OR \"cal... | No | 7.02 | 05-10s | ||
sample-005 | example-03 | conversation.material_request | Business and conversation | Does the caller request, accept or decline the offered information? | The agent can send programme information after the caller agrees to receive it. The agent is the AI assistant; the caller is the human/user. Use only this clip and the supplied context through the clip's end. | [
{
"id": "requested",
"description": "Requested: The caller initiates a request to receive material/information."
},
{
"id": "accepted",
"description": "Accepted: The caller accepts the agent's specific delivery offer."
},
{
"id": "declined",
"description": "Declined: The caller rejec... | accepted | 9.16 | 05-10s | ||
sample-006 | example-04 | conversation.response_to_question | Business and conversation | How does the caller's reply relate to the agent's pending question or statement? | The agent is wrapping up after learning that the caller has already joined a college. Listen to the caller's response at the end. The agent is the AI assistant; the caller is the human/user. Use only this clip and the supplied context through the clip's end. | [
{
"id": "answers_question",
"description": "Answers question: The reply supplies a usable answer to the actual pending question."
},
{
"id": "acknowledgment_only",
"description": "Acknowledgment only: It acknowledges hearing the agent without supplying the requested fact."
},
{
"id": "co... | contradicts_or_corrects | 7.3 | 05-10s | ||
sample-007 | example-06 | conversation.response_to_question | Business and conversation | How does the caller's reply relate to the agent's pending question or statement? | Listen only to the language used within this clip. A change within one speaker also counts. The agent is the AI assistant; the caller is the human/user. Use only this clip and the supplied context through the clip's end. | [
{
"id": "answers_question",
"description": "Answers question: The reply supplies a usable answer to the actual pending question."
},
{
"id": "acknowledgment_only",
"description": "Acknowledgment only: It acknowledges hearing the agent without supplying the requested fact."
},
{
"id": "co... | requires_clarification | 11.3 | 10-15s | ||
sample-008 | example-02 | graph.route_action | Graph routing | What should the agent do immediately after this clip? | Current node: opening. The agent is checking whether the caller has time for the conversation. The agent is the AI assistant; the caller is the human/user. Use only this clip and the supplied context through the clip's end.
Benchmark graph (reconstructed specification). The current node describes the clip start; track... | [
{
"id": "stay_or_clarify",
"description": "Preferred over wait whenever continuation is uncertain or both choices are plausible. Remain at the active stage for missing information, an on-topic clarification or audibility/connection repair. A completed questioning Hello?, Can you hear me?, I couldn't hear yo... | advance | opening | 7.02 | 05-10s | |
sample-009 | example-02 | graph.destination_node | Graph routing | Which graph node should the agent route to immediately after this clip? | Current node: opening. The agent is checking whether the caller has time for the conversation. The agent is the AI assistant; the caller is the human/user. Use only this clip and the supplied context through the clip's end.
Benchmark graph (reconstructed specification). The current node describes the clip start; track... | [
{
"id": "opening",
"description": "Route to this stage: Greet; establish the intended person or relative/proxy and availability to talk. Once established, move to interest."
},
{
"id": "interest",
"description": "Route to this stage: Establish admission interest, intended course/program and requ... | interest | opening | 7.02 | 05-10s | |
sample-010 | example-04 | graph.route_action | Graph routing | What should the agent do immediately after this clip? | The agent is wrapping up after learning that the caller has already joined a college. Listen to the caller's response at the end. The agent is the AI assistant; the caller is the human/user. Use only this clip and the supplied context through the clip's end.
Benchmark graph (reconstructed specification). The current n... | [
{
"id": "stay_or_clarify",
"description": "Preferred over wait whenever continuation is uncertain or both choices are plausible. Remain at the active stage for missing information, an on-topic clarification or audibility/connection repair. A completed questioning Hello?, Can you hear me?, I couldn't hear yo... | stay_or_clarify | opening | 7.3 | 05-10s | |
sample-011 | example-06 | graph.destination_node | Graph routing | Which graph node should the agent route to immediately after this clip? | Listen only to the language used within this clip. A change within one speaker also counts. The agent is the AI assistant; the caller is the human/user. Use only this clip and the supplied context through the clip's end.
Benchmark graph (reconstructed specification). The current node describes the clip start; track ch... | [
{
"id": "opening",
"description": "Route to this stage: Greet; establish the intended person or relative/proxy and availability to talk. Once established, move to interest."
},
{
"id": "interest",
"description": "Route to this stage: Establish admission interest, intended course/program and requ... | profile | profile | 11.3 | 10-15s | |
sample-012 | example-03 | external_tool.selection | Tool decisions | Which available external tool, if any, should the agent select next? | The agent can send programme information after the caller agrees to receive it. The agent is the AI assistant; the caller is the human/user. Use only this clip and the supplied context through the clip's end.
Available external tools:
[{"name": "create_counsellor_handoff", "description": "Create or assign a follow-up ... | [
{
"id": "create_counsellor_handoff",
"description": "Create counsellor handoff: Create or assign a follow-up task for an admissions counsellor in the configured CRM or support queue. Preconditions: A requested or accepted counsellor follow-up is established. The configured destination/team and the relevant ... | send_material | 9.16 | 05-10s | ||
sample-013 | example-04 | conversation.next_speaker | Turn taking | Who should speak or continue speaking immediately after this clip? | The agent is wrapping up after learning that the caller has already joined a college. Listen to the caller's response at the end. The agent is the AI assistant; the caller is the human/user. Use only this clip and the supplied context through the clip's end. | [
{
"id": "agent",
"description": "The caller finished their answer/question and the AI should respond, or the AI has an unfinished turn. Also covers a caller's audibility check or clear nonresponse requiring an AI re-prompt."
},
{
"id": "caller",
"description": "The AI finished a question requiri... | caller | 7.3 | 05-10s | ||
sample-014 | example-02 | conversation.next_speaker | Turn taking | Who should speak or continue speaking immediately after this clip? | Current node: opening. The agent is checking whether the caller has time for the conversation. The agent is the AI assistant; the caller is the human/user. Use only this clip and the supplied context through the clip's end. | [
{
"id": "agent",
"description": "The caller finished their answer/question and the AI should respond, or the AI has an unfinished turn. Also covers a caller's audibility check or clear nonresponse requiring an AI re-prompt."
},
{
"id": "caller",
"description": "The AI finished a question requiri... | agent | 7.02 | 05-10s | ||
sample-015 | example-06 | conversation.next_speaker | Turn taking | Who should speak or continue speaking immediately after this clip? | Listen only to the language used within this clip. A change within one speaker also counts. The agent is the AI assistant; the caller is the human/user. Use only this clip and the supplied context through the clip's end. | [
{
"id": "agent",
"description": "The caller finished their answer/question and the AI should respond, or the AI has an unfinished turn. Also covers a caller's audibility check or clear nonresponse requiring an AI re-prompt."
},
{
"id": "caller",
"description": "The AI finished a question requiri... | agent | 11.3 | 10-15s | ||
sample-016 | example-04 | acoustic.customer_emotion | Voice properties | What emotion or engagement does the caller predominantly express in this clip? | Judge only what is audible within this exact clip. The agent is the AI assistant; the caller is the person speaking with it. Do not infer the caller from anonymous STT speaker identifiers or use later conversation. | [
{
"id": "neutral_or_matter_of_fact",
"description": "Neutral / matter-of-fact: An ordinary, even delivery without a distinct emotional expression. Neutral is a valid answer, not missing evidence."
},
{
"id": "satisfied_or_pleased",
"description": "Satisfied / pleased: Audible contentment, relief... | confused_or_puzzled | 7.3 | 05-10s | ||
sample-017 | example-05 | acoustic.customer_emotion | Voice properties | What emotion or engagement does the caller predominantly express in this clip? | Judge only what is audible within this exact clip. The agent is the AI assistant; the caller is the person speaking with it. Do not infer the caller from anonymous STT speaker identifiers or use later conversation. | [
{
"id": "neutral_or_matter_of_fact",
"description": "Neutral / matter-of-fact: An ordinary, even delivery without a distinct emotional expression. Neutral is a valid answer, not missing evidence."
},
{
"id": "satisfied_or_pleased",
"description": "Satisfied / pleased: Audible contentment, relief... | satisfied_or_pleased | 7.5 | 05-10s | ||
sample-018 | example-04 | acoustic.speaker_count | Voice properties | How many perceptually distinct speakers are audible in this clip? | Judge only what is audible within this exact clip. The agent is the AI assistant; the caller is the person speaking with it. Do not infer the caller from anonymous STT speaker identifiers or use later conversation. | [
{
"id": "one",
"description": "One: One perceptually distinct voice is audible, whether agent or customer."
},
{
"id": "two",
"description": "Two: Two distinct voices are audible, even if speaking sequentially."
},
{
"id": "three_or_more",
"description": "Three or more: At least thre... | two | 7.3 | 05-10s | ||
sample-019 | example-02 | acoustic.disturbance | Voice properties | How much audible disturbance or interference is present in this clip? | Judge only what is audible within this exact clip. The agent is the AI assistant; the caller is the person speaking with it. Do not infer the caller from anonymous STT speaker identifiers or use later conversation. | [
{
"id": "none_or_minor",
"description": "None or minor: No notable interference, or minor noise without meaningful intelligibility loss."
},
{
"id": "noticeable_but_intelligible",
"description": "Noticeable but intelligible: Noise/interference is clearly audible but relevant speech remains intel... | none_or_minor | 7.02 | 05-10s | ||
sample-020 | example-06 | acoustic.language_switch | Language switching | Is there a language switch within this clip, within a speaker's speech or between speakers? | Judge only what is audible within this exact clip. The agent is the AI assistant; the caller is the person speaking with it. Do not infer the caller from anonymous STT speaker identifiers or use later conversation. | [
{
"id": "switch_detected",
"description": "Language switch detected: Two or more languages are audibly used within this clip, with a transition within one speaker's speech or between speakers. An English agent question followed by a Hindi caller reply counts."
},
{
"id": "no_switch_detected",
"d... | switch_detected | 11.3 | 10-15s |
DECIDE: 20 public samples
Decision Evaluation in Calls: Intent, Dialogue and Emotion
Website · GitHub · Full benchmark evaluation
This public preview contains exactly 20 labelled decision questions, using six audio clips already published on the DECIDE website. A clip can support multiple different questions. The examples span 14 task types across all six decision families. Audio comes from production voice-agent calls supplied by Bolna.
The full DECIDE v1.0 benchmark contains 4,867 questions over 1,081 clips and 32 tasks and remains private. This repository is a small, deliberately selected demonstration of question types, not the full benchmark or a representative test set. Use the hosted evaluator to obtain full-benchmark results.
Contents
| Decision family | Rows |
|---|---|
| Business and conversation | 7 |
| Graph routing | 4 |
| Language switching | 1 |
| Tool decisions | 1 |
| Turn taking | 3 |
| Voice properties | 4 |
| Total | 20 |
Examples include lead qualification, caller availability and callback requests, information delivery, caller responses, graph routing and next-node selection, external-tool selection, next-turn prediction, caller emotion, speaker count, audio disturbance and language switching.
Each row contains:
audio: a playable mono 16 kHz PCM16 WAV clip, embedded in the Parquet file.question,context,options: the decision question, permitted context and allowed option IDs with descriptions.answer: the existing human reference label from the frozen DECIDE v1.0 data.task,family: the task identifier and decision family.id,clip_id: public sample identifiers. Repeatedclip_idvalues mean the same audio is being used for different questions.current_node: the node at the clip start for graph questions; empty for other questions. The graph specification is included in that row's context.duration_seconds,duration_bucket: audio duration and duration group.
Load the preview
from datasets import Audio, load_dataset
samples = load_dataset("tazolabs/DECIDE-samples", split="preview")
assert len(samples) == 20
# Read encoded WAV bytes without installing an audio decoder.
samples = samples.cast_column("audio", Audio(decode=False))
row = samples[0]
print(row["question"])
print(row["options"])
print(row["answer"])
wav_bytes = row["audio"]["bytes"]
The Hugging Face dataset viewer provides an audio player for each row. For local
decoded-audio playback, use the audio dependencies supported by your version of
datasets.
Interpretation
The task is to choose one allowed decision at the end of the supplied clip.
Use only the audio and permitted context. Future conversation is not provided.
agent means the AI assistant; caller means the human/user. Graphs and tools are
reconstructed benchmark specifications, not production execution logs. Choosing
a tool does not execute it.
Reference answers are copied unchanged from existing human annotations. Context and question wording are shortened for public inspection; full-call histories, private source identifiers, reviewer notes and model predictions are excluded. Consequently, preview scores must not be reported as full DECIDE benchmark scores.
Only these 20 question rows are published here. No new call recordings beyond the six existing public website previews are released. The code repository's license does not grant a license to the underlying call recordings; no separate open-data license is specified for this preview.
Attribution
DECIDE is created by Tazo Labs using production call recordings supplied by Bolna. Authors: Harsh Pare, Naman Jain, Sachin Kumar Singh, Jayakrishna Sahit, Siddharth Batra, Sanket Bhat and Surya Vamsee Ayyagari. Paper/arXiv link: forthcoming. See the website for updates.
- Downloads last month
- 33