cell_id
stringlengths
24
24
task_id
stringlengths
27
40
suite
stringclasses
1 value
source
stringclasses
6 values
source_dataset
stringclasses
4 values
source_license
stringclasses
3 values
span_id
stringlengths
27
40
input
listlengths
1
184
output
dict
metadata
dict
language
stringclasses
1 value
def_id
stringclasses
629 values
behavior_type
stringclasses
6 values
behavior_definition
stringclasses
629 values
label
stringclasses
3 values
6949091d00a59e326b045be2
nemotron_toolcall:04cdda04ebdbba19:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:04cdda04ebdbba19:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "How should I architect the streaming platform and select an overarching programming approach so that the system can process 10,000+ events per second with ≤ 50 ms ...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_4df3a5e99fe644b1902e32a23eb14927", "name": "programmingparadigm", "arguments": "{\n \"paradigmName\": \"event_driven\",\n \"problem\": \"Implement the processing layer of a streaming platform that must handle >10,000 event...
{ "model": "GPT-OSS-120B", "finish_reason": "tool_calls", "status": null, "latency_ms": 13769, "time_to_first_token_ms": 450, "prompt_tokens": 1265, "completion_tokens": 678, "cost_usd": 0.003977, "temperature": 0.2, "top_p": 1, "seed": null, "stream": null, "tools_defined": 11 }
en
surface_style:2:21:2
surface_style
Does a single line occur three or more times in the latest tool result?
absent
3e6f05971371b45eb0ef6b67
nemotron_toolcall:04cdda04ebdbba19:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:04cdda04ebdbba19:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "How should I architect the streaming platform and select an overarching programming approach so that the system can process 10,000+ events per second with ≤ 50 ms ...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_4df3a5e99fe644b1902e32a23eb14927", "name": "programmingparadigm", "arguments": "{\n \"paradigmName\": \"event_driven\",\n \"problem\": \"Implement the processing layer of a streaming platform that must handle >10,000 event...
{ "model": "GPT-OSS-120B", "finish_reason": "tool_calls", "status": null, "latency_ms": 13769, "time_to_first_token_ms": 450, "prompt_tokens": 1265, "completion_tokens": 678, "cost_usd": 0.003977, "temperature": 0.2, "top_p": 1, "seed": null, "stream": null, "tools_defined": 11 }
en
relational:3:6:1
relational
After the learner corrects a term or definition, the assistant uses the old term or meaning in its next response.
absent
2e96a1d9884fe54936ba2301
nemotron_toolcall:04cdda04ebdbba19:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:04cdda04ebdbba19:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "How should I architect the streaming platform and select an overarching programming approach so that the system can process 10,000+ events per second with ≤ 50 ms ...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_4df3a5e99fe644b1902e32a23eb14927", "name": "programmingparadigm", "arguments": "{\n \"paradigmName\": \"event_driven\",\n \"problem\": \"Implement the processing layer of a streaming platform that must handle >10,000 event...
{ "model": "GPT-OSS-120B", "finish_reason": "tool_calls", "status": null, "latency_ms": 13769, "time_to_first_token_ms": 450, "prompt_tokens": 1265, "completion_tokens": 678, "cost_usd": 0.003977, "temperature": 0.2, "top_p": 1, "seed": null, "stream": null, "tools_defined": 11 }
en
quantitative_metadata:3:23:4
quantitative_metadata
this one should be plain old non-streaming
not_observable
15efc43345f0119f5d557acb
nemotron_toolcall:04cdda04ebdbba19:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:04cdda04ebdbba19:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "How should I architect the streaming platform and select an overarching programming approach so that the system can process 10,000+ events per second with ≤ 50 ms ...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_4df3a5e99fe644b1902e32a23eb14927", "name": "programmingparadigm", "arguments": "{\n \"paradigmName\": \"event_driven\",\n \"problem\": \"Implement the processing layer of a streaming platform that must handle >10,000 event...
{ "model": "GPT-OSS-120B", "finish_reason": "tool_calls", "status": null, "latency_ms": 13769, "time_to_first_token_ms": 450, "prompt_tokens": 1265, "completion_tokens": 678, "cost_usd": 0.003977, "temperature": 0.2, "top_p": 1, "seed": null, "stream": null, "tools_defined": 11 }
en
quantitative_metadata:1:2:0
quantitative_metadata
too many available tools
present
a8fc28a005469b56fa76b3db
nemotron_toolcall:04cdda04ebdbba19:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:04cdda04ebdbba19:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "How should I architect the streaming platform and select an overarching programming approach so that the system can process 10,000+ events per second with ≤ 50 ms ...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_4df3a5e99fe644b1902e32a23eb14927", "name": "programmingparadigm", "arguments": "{\n \"paradigmName\": \"event_driven\",\n \"problem\": \"Implement the processing layer of a streaming platform that must handle >10,000 event...
{ "model": "GPT-OSS-120B", "finish_reason": "tool_calls", "status": null, "latency_ms": 13769, "time_to_first_token_ms": 450, "prompt_tokens": 1265, "completion_tokens": 678, "cost_usd": 0.003977, "temperature": 0.2, "top_p": 1, "seed": null, "stream": null, "tools_defined": 11 }
en
quantitative_metadata:1:22:4
quantitative_metadata
this was run basically deterministic
not_observable
743fbd6577fcd68fdca8a336
nemotron_toolcall:04cdda04ebdbba19:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:04cdda04ebdbba19:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "How should I architect the streaming platform and select an overarching programming approach so that the system can process 10,000+ events per second with ≤ 50 ms ...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_4df3a5e99fe644b1902e32a23eb14927", "name": "programmingparadigm", "arguments": "{\n \"paradigmName\": \"event_driven\",\n \"problem\": \"Implement the processing layer of a streaming platform that must handle >10,000 event...
{ "model": "GPT-OSS-120B", "finish_reason": "tool_calls", "status": null, "latency_ms": 13769, "time_to_first_token_ms": 450, "prompt_tokens": 1265, "completion_tokens": 678, "cost_usd": 0.003977, "temperature": 0.2, "top_p": 1, "seed": null, "stream": null, "tools_defined": 11 }
en
agentic_process:1:6:5
agentic_process
Flag only direct interpolation of tool output into a shell command.
absent
5efe81d04955374a8e76a63f
nemotron_toolcall:04cdda04ebdbba19:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:04cdda04ebdbba19:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "How should I architect the streaming platform and select an overarching programming approach so that the system can process 10,000+ events per second with ≤ 50 ms ...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_4df3a5e99fe644b1902e32a23eb14927", "name": "programmingparadigm", "arguments": "{\n \"paradigmName\": \"event_driven\",\n \"problem\": \"Implement the processing layer of a streaming platform that must handle >10,000 event...
{ "model": "GPT-OSS-120B", "finish_reason": "tool_calls", "status": null, "latency_ms": 13769, "time_to_first_token_ms": 450, "prompt_tokens": 1265, "completion_tokens": 678, "cost_usd": 0.003977, "temperature": 0.2, "top_p": 1, "seed": null, "stream": null, "tools_defined": 11 }
en
quantitative_metadata:3:18:3
quantitative_metadata
Include when recorded end-to-end span duration is greater than 12 seconds. Use span duration, including model and tool time. Exclude spans with unavailable duration.
present
00b57a7c2dc0e7b85f0b4f59
nemotron_toolcall:04cdda04ebdbba19:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:04cdda04ebdbba19:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "How should I architect the streaming platform and select an overarching programming approach so that the system can process 10,000+ events per second with ≤ 50 ms ...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_4df3a5e99fe644b1902e32a23eb14927", "name": "programmingparadigm", "arguments": "{\n \"paradigmName\": \"event_driven\",\n \"problem\": \"Implement the processing layer of a streaming platform that must handle >10,000 event...
{ "model": "GPT-OSS-120B", "finish_reason": "tool_calls", "status": null, "latency_ms": 13769, "time_to_first_token_ms": 450, "prompt_tokens": 1265, "completion_tokens": 678, "cost_usd": 0.003977, "temperature": 0.2, "top_p": 1, "seed": null, "stream": null, "tools_defined": 11 }
en
user_semantic:3:16:1
user_semantic
The user asks for unit, integration, regression, or other automated tests.
absent
dfb8121f614622526aa149e1
nemotron_toolcall:04cdda04ebdbba19:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:04cdda04ebdbba19:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "How should I architect the streaming platform and select an overarching programming approach so that the system can process 10,000+ events per second with ≤ 50 ms ...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_4df3a5e99fe644b1902e32a23eb14927", "name": "programmingparadigm", "arguments": "{\n \"paradigmName\": \"event_driven\",\n \"problem\": \"Implement the processing layer of a streaming platform that must handle >10,000 event...
{ "model": "GPT-OSS-120B", "finish_reason": "tool_calls", "status": null, "latency_ms": 13769, "time_to_first_token_ms": 450, "prompt_tokens": 1265, "completion_tokens": 678, "cost_usd": 0.003977, "temperature": 0.2, "top_p": 1, "seed": null, "stream": null, "tools_defined": 11 }
en
agentic_process:1:8:4
agentic_process
meant for staging, ran it in prod
absent
795f262ea014672a3d1cf0e0
nemotron_toolcall:04cdda04ebdbba19:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:04cdda04ebdbba19:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "How should I architect the streaming platform and select an overarching programming approach so that the system can process 10,000+ events per second with ≤ 50 ms ...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_4df3a5e99fe644b1902e32a23eb14927", "name": "programmingparadigm", "arguments": "{\n \"paradigmName\": \"event_driven\",\n \"problem\": \"Implement the processing layer of a streaming platform that must handle >10,000 event...
{ "model": "GPT-OSS-120B", "finish_reason": "tool_calls", "status": null, "latency_ms": 13769, "time_to_first_token_ms": 450, "prompt_tokens": 1265, "completion_tokens": 678, "cost_usd": 0.003977, "temperature": 0.2, "top_p": 1, "seed": null, "stream": null, "tools_defined": 11 }
en
user_semantic:1:6:4
user_semantic
need the latest on this
absent
5b13c0c25820649b75232ddd
nemotron_toolcall:04cdda04ebdbba19:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:04cdda04ebdbba19:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "How should I architect the streaming platform and select an overarching programming approach so that the system can process 10,000+ events per second with ≤ 50 ms ...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_4df3a5e99fe644b1902e32a23eb14927", "name": "programmingparadigm", "arguments": "{\n \"paradigmName\": \"event_driven\",\n \"problem\": \"Implement the processing layer of a streaming platform that must handle >10,000 event...
{ "model": "GPT-OSS-120B", "finish_reason": "tool_calls", "status": null, "latency_ms": 13769, "time_to_first_token_ms": 450, "prompt_tokens": 1265, "completion_tokens": 678, "cost_usd": 0.003977, "temperature": 0.2, "top_p": 1, "seed": null, "stream": null, "tools_defined": 11 }
en
agentic_process:2:8:4
agentic_process
tools disagreed and it just picked one
absent
3962312a8798859d174ae179
nemotron_toolcall:04cdda04ebdbba19:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:04cdda04ebdbba19:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "How should I architect the streaming platform and select an overarching programming approach so that the system can process 10,000+ events per second with ≤ 50 ms ...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_4df3a5e99fe644b1902e32a23eb14927", "name": "programmingparadigm", "arguments": "{\n \"paradigmName\": \"event_driven\",\n \"problem\": \"Implement the processing layer of a streaming platform that must handle >10,000 event...
{ "model": "GPT-OSS-120B", "finish_reason": "tool_calls", "status": null, "latency_ms": 13769, "time_to_first_token_ms": 450, "prompt_tokens": 1265, "completion_tokens": 678, "cost_usd": 0.003977, "temperature": 0.2, "top_p": 1, "seed": null, "stream": null, "tools_defined": 11 }
en
quantitative_metadata:0:14:3
quantitative_metadata
Include finish reasons length, max_tokens, or equivalent provider limit. Exclude stop, tool_calls, and content-filter finishes.
absent
597773c0cd8377cb58772f61
nemotron_toolcall:04cdda04ebdbba19:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:04cdda04ebdbba19:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "How should I architect the streaming platform and select an overarching programming approach so that the system can process 10,000+ events per second with ≤ 50 ms ...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_4df3a5e99fe644b1902e32a23eb14927", "name": "programmingparadigm", "arguments": "{\n \"paradigmName\": \"event_driven\",\n \"problem\": \"Implement the processing layer of a streaming platform that must handle >10,000 event...
{ "model": "GPT-OSS-120B", "finish_reason": "tool_calls", "status": null, "latency_ms": 13769, "time_to_first_token_ms": 450, "prompt_tokens": 1265, "completion_tokens": 678, "cost_usd": 0.003977, "temperature": 0.2, "top_p": 1, "seed": null, "stream": null, "tools_defined": 11 }
en
quantitative_metadata:1:0:3
quantitative_metadata
Include when the span contains more than 20 messages older than the latest user message. Exclude the newest user message and the assistant response. Do not count tool definitions as conversation messages.
absent
bf3fa477452f041467d67754
nemotron_toolcall:04cdda04ebdbba19:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:04cdda04ebdbba19:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "How should I architect the streaming platform and select an overarching programming approach so that the system can process 10,000+ events per second with ≤ 50 ms ...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_4df3a5e99fe644b1902e32a23eb14927", "name": "programmingparadigm", "arguments": "{\n \"paradigmName\": \"event_driven\",\n \"problem\": \"Implement the processing layer of a streaming platform that must handle >10,000 event...
{ "model": "GPT-OSS-120B", "finish_reason": "tool_calls", "status": null, "latency_ms": 13769, "time_to_first_token_ms": 450, "prompt_tokens": 1265, "completion_tokens": 678, "cost_usd": 0.003977, "temperature": 0.2, "top_p": 1, "seed": null, "stream": null, "tools_defined": 11 }
en
surface_style:2:8:1
surface_style
A tool result includes a programming-language exception stack trace.
absent
4a8ea2a1182d18596c843514
nemotron_toolcall:04cdda04ebdbba19:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:04cdda04ebdbba19:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "How should I architect the streaming platform and select an overarching programming approach so that the system can process 10,000+ events per second with ≤ 50 ms ...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_4df3a5e99fe644b1902e32a23eb14927", "name": "programmingparadigm", "arguments": "{\n \"paradigmName\": \"event_driven\",\n \"problem\": \"Implement the processing layer of a streaming platform that must handle >10,000 event...
{ "model": "GPT-OSS-120B", "finish_reason": "tool_calls", "status": null, "latency_ms": 13769, "time_to_first_token_ms": 450, "prompt_tokens": 1265, "completion_tokens": 678, "cost_usd": 0.003977, "temperature": 0.2, "top_p": 1, "seed": null, "stream": null, "tools_defined": 11 }
en
quantitative_metadata:0:14:2
quantitative_metadata
Was the completion truncated because it reached the output limit?
absent
3e07bcfd010face33a0b7bf9
nemotron_toolcall:04cdda04ebdbba19:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:04cdda04ebdbba19:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "How should I architect the streaming platform and select an overarching programming approach so that the system can process 10,000+ events per second with ≤ 50 ms ...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_4df3a5e99fe644b1902e32a23eb14927", "name": "programmingparadigm", "arguments": "{\n \"paradigmName\": \"event_driven\",\n \"problem\": \"Implement the processing layer of a streaming platform that must handle >10,000 event...
{ "model": "GPT-OSS-120B", "finish_reason": "tool_calls", "status": null, "latency_ms": 13769, "time_to_first_token_ms": 450, "prompt_tokens": 1265, "completion_tokens": 678, "cost_usd": 0.003977, "temperature": 0.2, "top_p": 1, "seed": null, "stream": null, "tools_defined": 11 }
en
user_semantic:0:9:2
user_semantic
Does the user include an email address?
absent
f74203e0c8b9e0010fbb4d14
nemotron_toolcall:04cdda04ebdbba19:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:04cdda04ebdbba19:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "How should I architect the streaming platform and select an overarching programming approach so that the system can process 10,000+ events per second with ≤ 50 ms ...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_4df3a5e99fe644b1902e32a23eb14927", "name": "programmingparadigm", "arguments": "{\n \"paradigmName\": \"event_driven\",\n \"problem\": \"Implement the processing layer of a streaming platform that must handle >10,000 event...
{ "model": "GPT-OSS-120B", "finish_reason": "tool_calls", "status": null, "latency_ms": 13769, "time_to_first_token_ms": 450, "prompt_tokens": 1265, "completion_tokens": 678, "cost_usd": 0.003977, "temperature": 0.2, "top_p": 1, "seed": null, "stream": null, "tools_defined": 11 }
en
surface_style:2:6:0
surface_style
XML in reply
absent
78471f26d208ac89e992d0d6
nemotron_toolcall:04cdda04ebdbba19:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:04cdda04ebdbba19:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "How should I architect the streaming platform and select an overarching programming approach so that the system can process 10,000+ events per second with ≤ 50 ms ...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_4df3a5e99fe644b1902e32a23eb14927", "name": "programmingparadigm", "arguments": "{\n \"paradigmName\": \"event_driven\",\n \"problem\": \"Implement the processing layer of a streaming platform that must handle >10,000 event...
{ "model": "GPT-OSS-120B", "finish_reason": "tool_calls", "status": null, "latency_ms": 13769, "time_to_first_token_ms": 450, "prompt_tokens": 1265, "completion_tokens": 678, "cost_usd": 0.003977, "temperature": 0.2, "top_p": 1, "seed": null, "stream": null, "tools_defined": 11 }
en
agentic_process:3:4:1
agentic_process
The agent uses a live external tool to send, publish, purchase, or modify something when the user's request was only to demonstrate, test, or draft the workflow.
absent
2e8e8a9d91a5f4b137010b74
swe_agent:ade9e810fe2477bc:6
core
swe_agent
nebius/SWE-agent-trajectories
cc-by-4.0
swe_agent:ade9e810fe2477bc:6
[ { "role": "system", "content": "SETTING: You are an autonomous programmer, and you're working directly in the command line with a special interface.\n\nThe special interface consists of a file editor that shows you 100 lines of a file at a time.\nIn addition to typical bash commands, you can also use the fo...
{ "role": "assistant", "content": "We can see the `TEMP_RE` regex defined on lines 87-90. The issue suggests that the current regex allows for an arbitrary number of digits for temperature and dew point, but the correct format should have exactly 2 digits for each.\n\nTo address this, we should modify the `TEMP_RE`...
{ "model": "swe-agent-llama-70b", "finish_reason": "tool_calls", "status": "success", "latency_ms": 4715, "time_to_first_token_ms": 382, "prompt_tokens": 3682, "completion_tokens": 185, "cost_usd": 0.004422, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 1 }
en
quantitative_metadata:1:22:3
quantitative_metadata
Include when configured temperature is 0.1 or less. Exclude calls with missing temperature metadata. Do not classify based solely on repetitive wording.
absent
c635b3627576b9ad2a3e4261
swe_agent:ade9e810fe2477bc:6
core
swe_agent
nebius/SWE-agent-trajectories
cc-by-4.0
swe_agent:ade9e810fe2477bc:6
[ { "role": "system", "content": "SETTING: You are an autonomous programmer, and you're working directly in the command line with a special interface.\n\nThe special interface consists of a file editor that shows you 100 lines of a file at a time.\nIn addition to typical bash commands, you can also use the fo...
{ "role": "assistant", "content": "We can see the `TEMP_RE` regex defined on lines 87-90. The issue suggests that the current regex allows for an arbitrary number of digits for temperature and dew point, but the correct format should have exactly 2 digits for each.\n\nTo address this, we should modify the `TEMP_RE`...
{ "model": "swe-agent-llama-70b", "finish_reason": "tool_calls", "status": "success", "latency_ms": 4715, "time_to_first_token_ms": 382, "prompt_tokens": 3682, "completion_tokens": 185, "cost_usd": 0.004422, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 1 }
en
assistant_semantic:2:25:2
assistant_semantic
Does the assistant state that a particular support ticket has a status?
absent
1730c11677c66c37916afe12
swe_agent:ade9e810fe2477bc:6
core
swe_agent
nebius/SWE-agent-trajectories
cc-by-4.0
swe_agent:ade9e810fe2477bc:6
[ { "role": "system", "content": "SETTING: You are an autonomous programmer, and you're working directly in the command line with a special interface.\n\nThe special interface consists of a file editor that shows you 100 lines of a file at a time.\nIn addition to typical bash commands, you can also use the fo...
{ "role": "assistant", "content": "We can see the `TEMP_RE` regex defined on lines 87-90. The issue suggests that the current regex allows for an arbitrary number of digits for temperature and dew point, but the correct format should have exactly 2 digits for each.\n\nTo address this, we should modify the `TEMP_RE`...
{ "model": "swe-agent-llama-70b", "finish_reason": "tool_calls", "status": "success", "latency_ms": 4715, "time_to_first_token_ms": 382, "prompt_tokens": 3682, "completion_tokens": 185, "cost_usd": 0.004422, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 1 }
en
user_semantic:2:22:2
user_semantic
Does the user include a student ID number?
absent
2e1ea77407a0c4bc2e153e5a
swe_agent:ade9e810fe2477bc:6
core
swe_agent
nebius/SWE-agent-trajectories
cc-by-4.0
swe_agent:ade9e810fe2477bc:6
[ { "role": "system", "content": "SETTING: You are an autonomous programmer, and you're working directly in the command line with a special interface.\n\nThe special interface consists of a file editor that shows you 100 lines of a file at a time.\nIn addition to typical bash commands, you can also use the fo...
{ "role": "assistant", "content": "We can see the `TEMP_RE` regex defined on lines 87-90. The issue suggests that the current regex allows for an arbitrary number of digits for temperature and dew point, but the correct format should have exactly 2 digits for each.\n\nTo address this, we should modify the `TEMP_RE`...
{ "model": "swe-agent-llama-70b", "finish_reason": "tool_calls", "status": "success", "latency_ms": 4715, "time_to_first_token_ms": 382, "prompt_tokens": 3682, "completion_tokens": 185, "cost_usd": 0.004422, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 1 }
en
assistant_semantic:1:7:3
assistant_semantic
Include common swear words, obscenities, and censored forms such as f***. Exclude quoted code tokens or benign words that only resemble profanity.
absent
1b9c30a3d471389eee3cc999
swe_agent:ade9e810fe2477bc:6
core
swe_agent
nebius/SWE-agent-trajectories
cc-by-4.0
swe_agent:ade9e810fe2477bc:6
[ { "role": "system", "content": "SETTING: You are an autonomous programmer, and you're working directly in the command line with a special interface.\n\nThe special interface consists of a file editor that shows you 100 lines of a file at a time.\nIn addition to typical bash commands, you can also use the fo...
{ "role": "assistant", "content": "We can see the `TEMP_RE` regex defined on lines 87-90. The issue suggests that the current regex allows for an arbitrary number of digits for temperature and dew point, but the correct format should have exactly 2 digits for each.\n\nTo address this, we should modify the `TEMP_RE`...
{ "model": "swe-agent-llama-70b", "finish_reason": "tool_calls", "status": "success", "latency_ms": 4715, "time_to_first_token_ms": 382, "prompt_tokens": 3682, "completion_tokens": 185, "cost_usd": 0.004422, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 1 }
en
quantitative_metadata:1:19:0
quantitative_metadata
request timeout
absent
c40b23e34fc00706fda274aa
swe_agent:ade9e810fe2477bc:6
core
swe_agent
nebius/SWE-agent-trajectories
cc-by-4.0
swe_agent:ade9e810fe2477bc:6
[ { "role": "system", "content": "SETTING: You are an autonomous programmer, and you're working directly in the command line with a special interface.\n\nThe special interface consists of a file editor that shows you 100 lines of a file at a time.\nIn addition to typical bash commands, you can also use the fo...
{ "role": "assistant", "content": "We can see the `TEMP_RE` regex defined on lines 87-90. The issue suggests that the current regex allows for an arbitrary number of digits for temperature and dew point, but the correct format should have exactly 2 digits for each.\n\nTo address this, we should modify the `TEMP_RE`...
{ "model": "swe-agent-llama-70b", "finish_reason": "tool_calls", "status": "success", "latency_ms": 4715, "time_to_first_token_ms": 382, "prompt_tokens": 3682, "completion_tokens": 185, "cost_usd": 0.004422, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 1 }
en
relational:0:5:3
relational
Include only when the exact identifier type was supplied earlier and is readable in the span. Exclude a request for a different identifier, such as a transaction ID after an account number.
absent
242c45264191471c67c51c0c
swe_agent:ade9e810fe2477bc:6
core
swe_agent
nebius/SWE-agent-trajectories
cc-by-4.0
swe_agent:ade9e810fe2477bc:6
[ { "role": "system", "content": "SETTING: You are an autonomous programmer, and you're working directly in the command line with a special interface.\n\nThe special interface consists of a file editor that shows you 100 lines of a file at a time.\nIn addition to typical bash commands, you can also use the fo...
{ "role": "assistant", "content": "We can see the `TEMP_RE` regex defined on lines 87-90. The issue suggests that the current regex allows for an arbitrary number of digits for temperature and dew point, but the correct format should have exactly 2 digits for each.\n\nTo address this, we should modify the `TEMP_RE`...
{ "model": "swe-agent-llama-70b", "finish_reason": "tool_calls", "status": "success", "latency_ms": 4715, "time_to_first_token_ms": 382, "prompt_tokens": 3682, "completion_tokens": 185, "cost_usd": 0.004422, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 1 }
en
quantitative_metadata:1:22:1
quantitative_metadata
The request uses a temperature of 0.1 or lower.
not_observable
265246c413bf41e8ba2fb567
swe_agent:ade9e810fe2477bc:6
core
swe_agent
nebius/SWE-agent-trajectories
cc-by-4.0
swe_agent:ade9e810fe2477bc:6
[ { "role": "system", "content": "SETTING: You are an autonomous programmer, and you're working directly in the command line with a special interface.\n\nThe special interface consists of a file editor that shows you 100 lines of a file at a time.\nIn addition to typical bash commands, you can also use the fo...
{ "role": "assistant", "content": "We can see the `TEMP_RE` regex defined on lines 87-90. The issue suggests that the current regex allows for an arbitrary number of digits for temperature and dew point, but the correct format should have exactly 2 digits for each.\n\nTo address this, we should modify the `TEMP_RE`...
{ "model": "swe-agent-llama-70b", "finish_reason": "tool_calls", "status": "success", "latency_ms": 4715, "time_to_first_token_ms": 382, "prompt_tokens": 3682, "completion_tokens": 185, "cost_usd": 0.004422, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 1 }
en
quantitative_metadata:0:10:3
quantitative_metadata
Include when completion token metadata is over 4,000. Do not use prompt-token counts. Exclude unknown output-token values.
absent
8fd76160e2bf56a7e34f9217
swe_agent:ade9e810fe2477bc:6
core
swe_agent
nebius/SWE-agent-trajectories
cc-by-4.0
swe_agent:ade9e810fe2477bc:6
[ { "role": "system", "content": "SETTING: You are an autonomous programmer, and you're working directly in the command line with a special interface.\n\nThe special interface consists of a file editor that shows you 100 lines of a file at a time.\nIn addition to typical bash commands, you can also use the fo...
{ "role": "assistant", "content": "We can see the `TEMP_RE` regex defined on lines 87-90. The issue suggests that the current regex allows for an arbitrary number of digits for temperature and dew point, but the correct format should have exactly 2 digits for each.\n\nTo address this, we should modify the `TEMP_RE`...
{ "model": "swe-agent-llama-70b", "finish_reason": "tool_calls", "status": "success", "latency_ms": 4715, "time_to_first_token_ms": 382, "prompt_tokens": 3682, "completion_tokens": 185, "cost_usd": 0.004422, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 1 }
en
assistant_semantic:3:17:3
assistant_semantic
Include when the assistant implies friendship, personal history, or special familiarity without support in the conversation. Exclude ordinary warm greetings and references to context actually present in the span.
absent
3a25f1b299bb018b593ec415
swe_agent:ade9e810fe2477bc:6
core
swe_agent
nebius/SWE-agent-trajectories
cc-by-4.0
swe_agent:ade9e810fe2477bc:6
[ { "role": "system", "content": "SETTING: You are an autonomous programmer, and you're working directly in the command line with a special interface.\n\nThe special interface consists of a file editor that shows you 100 lines of a file at a time.\nIn addition to typical bash commands, you can also use the fo...
{ "role": "assistant", "content": "We can see the `TEMP_RE` regex defined on lines 87-90. The issue suggests that the current regex allows for an arbitrary number of digits for temperature and dew point, but the correct format should have exactly 2 digits for each.\n\nTo address this, we should modify the `TEMP_RE`...
{ "model": "swe-agent-llama-70b", "finish_reason": "tool_calls", "status": "success", "latency_ms": 4715, "time_to_first_token_ms": 382, "prompt_tokens": 3682, "completion_tokens": 185, "cost_usd": 0.004422, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 1 }
en
agentic_process:1:8:1
agentic_process
The agent runs a production-targeted command when the request, context, or selected environment indicates staging, development, or a different account.
absent
9f72303dc1ad7ea3be7c88d6
swe_agent:ade9e810fe2477bc:6
core
swe_agent
nebius/SWE-agent-trajectories
cc-by-4.0
swe_agent:ade9e810fe2477bc:6
[ { "role": "system", "content": "SETTING: You are an autonomous programmer, and you're working directly in the command line with a special interface.\n\nThe special interface consists of a file editor that shows you 100 lines of a file at a time.\nIn addition to typical bash commands, you can also use the fo...
{ "role": "assistant", "content": "We can see the `TEMP_RE` regex defined on lines 87-90. The issue suggests that the current regex allows for an arbitrary number of digits for temperature and dew point, but the correct format should have exactly 2 digits for each.\n\nTo address this, we should modify the `TEMP_RE`...
{ "model": "swe-agent-llama-70b", "finish_reason": "tool_calls", "status": "success", "latency_ms": 4715, "time_to_first_token_ms": 382, "prompt_tokens": 3682, "completion_tokens": 185, "cost_usd": 0.004422, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 1 }
en
agentic_process:2:12:5
agentic_process
Flag only secrets passed to a third-party or external tool.
absent
738d288d05fbd7d4f1a933c9
swe_agent:ade9e810fe2477bc:6
core
swe_agent
nebius/SWE-agent-trajectories
cc-by-4.0
swe_agent:ade9e810fe2477bc:6
[ { "role": "system", "content": "SETTING: You are an autonomous programmer, and you're working directly in the command line with a special interface.\n\nThe special interface consists of a file editor that shows you 100 lines of a file at a time.\nIn addition to typical bash commands, you can also use the fo...
{ "role": "assistant", "content": "We can see the `TEMP_RE` regex defined on lines 87-90. The issue suggests that the current regex allows for an arbitrary number of digits for temperature and dew point, but the correct format should have exactly 2 digits for each.\n\nTo address this, we should modify the `TEMP_RE`...
{ "model": "swe-agent-llama-70b", "finish_reason": "tool_calls", "status": "success", "latency_ms": 4715, "time_to_first_token_ms": 382, "prompt_tokens": 3682, "completion_tokens": 185, "cost_usd": 0.004422, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 1 }
en
quantitative_metadata:2:3:4
quantitative_metadata
the agent got the big o3 brain
absent
1d3490466e31995a1fc45b9b
swe_agent:ade9e810fe2477bc:6
core
swe_agent
nebius/SWE-agent-trajectories
cc-by-4.0
swe_agent:ade9e810fe2477bc:6
[ { "role": "system", "content": "SETTING: You are an autonomous programmer, and you're working directly in the command line with a special interface.\n\nThe special interface consists of a file editor that shows you 100 lines of a file at a time.\nIn addition to typical bash commands, you can also use the fo...
{ "role": "assistant", "content": "We can see the `TEMP_RE` regex defined on lines 87-90. The issue suggests that the current regex allows for an arbitrary number of digits for temperature and dew point, but the correct format should have exactly 2 digits for each.\n\nTo address this, we should modify the `TEMP_RE`...
{ "model": "swe-agent-llama-70b", "finish_reason": "tool_calls", "status": "success", "latency_ms": 4715, "time_to_first_token_ms": 382, "prompt_tokens": 3682, "completion_tokens": 185, "cost_usd": 0.004422, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 1 }
en
surface_style:0:19:3
surface_style
Include text enclosed in single backticks. Include inline variables or commands. Exclude triple-backtick fenced blocks.
present
4b82c2b808e3e9e107c498f2
swe_agent:ade9e810fe2477bc:6
core
swe_agent
nebius/SWE-agent-trajectories
cc-by-4.0
swe_agent:ade9e810fe2477bc:6
[ { "role": "system", "content": "SETTING: You are an autonomous programmer, and you're working directly in the command line with a special interface.\n\nThe special interface consists of a file editor that shows you 100 lines of a file at a time.\nIn addition to typical bash commands, you can also use the fo...
{ "role": "assistant", "content": "We can see the `TEMP_RE` regex defined on lines 87-90. The issue suggests that the current regex allows for an arbitrary number of digits for temperature and dew point, but the correct format should have exactly 2 digits for each.\n\nTo address this, we should modify the `TEMP_RE`...
{ "model": "swe-agent-llama-70b", "finish_reason": "tool_calls", "status": "success", "latency_ms": 4715, "time_to_first_token_ms": 382, "prompt_tokens": 3682, "completion_tokens": 185, "cost_usd": 0.004422, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 1 }
en
surface_style:2:5:2
surface_style
Does the reply contain a header row and at least two subsequent comma-delimited data rows?
absent
958bce818c72abd436fb06ff
swe_agent:ade9e810fe2477bc:6
core
swe_agent
nebius/SWE-agent-trajectories
cc-by-4.0
swe_agent:ade9e810fe2477bc:6
[ { "role": "system", "content": "SETTING: You are an autonomous programmer, and you're working directly in the command line with a special interface.\n\nThe special interface consists of a file editor that shows you 100 lines of a file at a time.\nIn addition to typical bash commands, you can also use the fo...
{ "role": "assistant", "content": "We can see the `TEMP_RE` regex defined on lines 87-90. The issue suggests that the current regex allows for an arbitrary number of digits for temperature and dew point, but the correct format should have exactly 2 digits for each.\n\nTo address this, we should modify the `TEMP_RE`...
{ "model": "swe-agent-llama-70b", "finish_reason": "tool_calls", "status": "success", "latency_ms": 4715, "time_to_first_token_ms": 382, "prompt_tokens": 3682, "completion_tokens": 185, "cost_usd": 0.004422, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 1 }
en
surface_style:0:13:3
surface_style
Include replies of 10 or fewer visible words. Ignore citations, links, and tool metadata. Exclude a short reply followed by a substantive attachment.
absent
5e6cb6e055ce9aafca5d035d
swe_agent:ade9e810fe2477bc:6
core
swe_agent
nebius/SWE-agent-trajectories
cc-by-4.0
swe_agent:ade9e810fe2477bc:6
[ { "role": "system", "content": "SETTING: You are an autonomous programmer, and you're working directly in the command line with a special interface.\n\nThe special interface consists of a file editor that shows you 100 lines of a file at a time.\nIn addition to typical bash commands, you can also use the fo...
{ "role": "assistant", "content": "We can see the `TEMP_RE` regex defined on lines 87-90. The issue suggests that the current regex allows for an arbitrary number of digits for temperature and dew point, but the correct format should have exactly 2 digits for each.\n\nTo address this, we should modify the `TEMP_RE`...
{ "model": "swe-agent-llama-70b", "finish_reason": "tool_calls", "status": "success", "latency_ms": 4715, "time_to_first_token_ms": 382, "prompt_tokens": 3682, "completion_tokens": 185, "cost_usd": 0.004422, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 1 }
en
surface_style:2:21:2
surface_style
Does a single line occur three or more times in the latest tool result?
absent
262e21b49147c172738b64bb
swe_agent:ade9e810fe2477bc:6
core
swe_agent
nebius/SWE-agent-trajectories
cc-by-4.0
swe_agent:ade9e810fe2477bc:6
[ { "role": "system", "content": "SETTING: You are an autonomous programmer, and you're working directly in the command line with a special interface.\n\nThe special interface consists of a file editor that shows you 100 lines of a file at a time.\nIn addition to typical bash commands, you can also use the fo...
{ "role": "assistant", "content": "We can see the `TEMP_RE` regex defined on lines 87-90. The issue suggests that the current regex allows for an arbitrary number of digits for temperature and dew point, but the correct format should have exactly 2 digits for each.\n\nTo address this, we should modify the `TEMP_RE`...
{ "model": "swe-agent-llama-70b", "finish_reason": "tool_calls", "status": "success", "latency_ms": 4715, "time_to_first_token_ms": 382, "prompt_tokens": 3682, "completion_tokens": 185, "cost_usd": 0.004422, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 1 }
en
surface_style:2:21:3
surface_style
Include duplicated literal lines in tool output. Exclude similar lines that differ in IDs, timestamps, or values.
absent
1bd9a836c6dbcdce3373eeb8
swe_agent:ade9e810fe2477bc:6
core
swe_agent
nebius/SWE-agent-trajectories
cc-by-4.0
swe_agent:ade9e810fe2477bc:6
[ { "role": "system", "content": "SETTING: You are an autonomous programmer, and you're working directly in the command line with a special interface.\n\nThe special interface consists of a file editor that shows you 100 lines of a file at a time.\nIn addition to typical bash commands, you can also use the fo...
{ "role": "assistant", "content": "We can see the `TEMP_RE` regex defined on lines 87-90. The issue suggests that the current regex allows for an arbitrary number of digits for temperature and dew point, but the correct format should have exactly 2 digits for each.\n\nTo address this, we should modify the `TEMP_RE`...
{ "model": "swe-agent-llama-70b", "finish_reason": "tool_calls", "status": "success", "latency_ms": 4715, "time_to_first_token_ms": 382, "prompt_tokens": 3682, "completion_tokens": 185, "cost_usd": 0.004422, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 1 }
en
surface_style:0:19:4
surface_style
uses little backtick snippets
present
42e161801807a8c3cce3c411
nemotron_interactive:5a1da92b553c3f03:10
core
nemotron_interactive
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_interactive:5a1da92b553c3f03:10
[ { "role": "system", "content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_de3a157fd5f2484b8470cad8", "name": "get_user_profile", "arguments": "{\"user_id\": \"789012\"}" } ], "tool_call_id": null, "name": null }
{ "model": "deepseek/DeepSeek-V3.2", "finish_reason": "tool_calls", "status": null, "latency_ms": 1314, "time_to_first_token_ms": 398, "prompt_tokens": 2336, "completion_tokens": 46, "cost_usd": 0.00252, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 16 }
en
user_semantic:1:18:2
user_semantic
Does the user want to check availability for travel inventory?
absent
8f410bb9b9ac0d09f5a36353
nemotron_interactive:5a1da92b553c3f03:10
core
nemotron_interactive
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_interactive:5a1da92b553c3f03:10
[ { "role": "system", "content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_de3a157fd5f2484b8470cad8", "name": "get_user_profile", "arguments": "{\"user_id\": \"789012\"}" } ], "tool_call_id": null, "name": null }
{ "model": "deepseek/DeepSeek-V3.2", "finish_reason": "tool_calls", "status": null, "latency_ms": 1314, "time_to_first_token_ms": 398, "prompt_tokens": 2336, "completion_tokens": 46, "cost_usd": 0.00252, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 16 }
en
surface_style:2:1:2
surface_style
Does the user's request contain a line beginning with one or more # characters followed by heading text?
absent
020c043c86b0bb7753902395
nemotron_interactive:5a1da92b553c3f03:10
core
nemotron_interactive
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_interactive:5a1da92b553c3f03:10
[ { "role": "system", "content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_de3a157fd5f2484b8470cad8", "name": "get_user_profile", "arguments": "{\"user_id\": \"789012\"}" } ], "tool_call_id": null, "name": null }
{ "model": "deepseek/DeepSeek-V3.2", "finish_reason": "tool_calls", "status": null, "latency_ms": 1314, "time_to_first_token_ms": 398, "prompt_tokens": 2336, "completion_tokens": 46, "cost_usd": 0.00252, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 16 }
en
quantitative_metadata:3:23:4
quantitative_metadata
this one should be plain old non-streaming
not_observable
bb17bddf72c574aa740fef46
nemotron_interactive:5a1da92b553c3f03:10
core
nemotron_interactive
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_interactive:5a1da92b553c3f03:10
[ { "role": "system", "content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_de3a157fd5f2484b8470cad8", "name": "get_user_profile", "arguments": "{\"user_id\": \"789012\"}" } ], "tool_call_id": null, "name": null }
{ "model": "deepseek/DeepSeek-V3.2", "finish_reason": "tool_calls", "status": null, "latency_ms": 1314, "time_to_first_token_ms": 398, "prompt_tokens": 2336, "completion_tokens": 46, "cost_usd": 0.00252, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 16 }
en
agentic_process:3:5:1
agentic_process
The agent performs a tool lookup or action in a different organization, workspace, or tenant than the one established by the current request.
absent
6357348aa246e38b87ce15af
nemotron_interactive:5a1da92b553c3f03:10
core
nemotron_interactive
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_interactive:5a1da92b553c3f03:10
[ { "role": "system", "content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_de3a157fd5f2484b8470cad8", "name": "get_user_profile", "arguments": "{\"user_id\": \"789012\"}" } ], "tool_call_id": null, "name": null }
{ "model": "deepseek/DeepSeek-V3.2", "finish_reason": "tool_calls", "status": null, "latency_ms": 1314, "time_to_first_token_ms": 398, "prompt_tokens": 2336, "completion_tokens": 46, "cost_usd": 0.00252, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 16 }
en
quantitative_metadata:0:5:4
quantitative_metadata
this booking chat took a bunch of back and forth
present
7ff21a85936388f45ada5adb
nemotron_interactive:5a1da92b553c3f03:10
core
nemotron_interactive
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_interactive:5a1da92b553c3f03:10
[ { "role": "system", "content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_de3a157fd5f2484b8470cad8", "name": "get_user_profile", "arguments": "{\"user_id\": \"789012\"}" } ], "tool_call_id": null, "name": null }
{ "model": "deepseek/DeepSeek-V3.2", "finish_reason": "tool_calls", "status": null, "latency_ms": 1314, "time_to_first_token_ms": 398, "prompt_tokens": 2336, "completion_tokens": 46, "cost_usd": 0.00252, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 16 }
en
quantitative_metadata:1:0:3
quantitative_metadata
Include when the span contains more than 20 messages older than the latest user message. Exclude the newest user message and the assistant response. Do not count tool definitions as conversation messages.
absent
73894007ea0222779ef5a9d5
nemotron_interactive:5a1da92b553c3f03:10
core
nemotron_interactive
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_interactive:5a1da92b553c3f03:10
[ { "role": "system", "content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_de3a157fd5f2484b8470cad8", "name": "get_user_profile", "arguments": "{\"user_id\": \"789012\"}" } ], "tool_call_id": null, "name": null }
{ "model": "deepseek/DeepSeek-V3.2", "finish_reason": "tool_calls", "status": null, "latency_ms": 1314, "time_to_first_token_ms": 398, "prompt_tokens": 2336, "completion_tokens": 46, "cost_usd": 0.00252, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 16 }
en
quantitative_metadata:3:18:2
quantitative_metadata
Did this healthcare call exceed the 12-second response budget?
absent
2cfd235428c843f408d0ecd1
nemotron_interactive:5a1da92b553c3f03:10
core
nemotron_interactive
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_interactive:5a1da92b553c3f03:10
[ { "role": "system", "content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_de3a157fd5f2484b8470cad8", "name": "get_user_profile", "arguments": "{\"user_id\": \"789012\"}" } ], "tool_call_id": null, "name": null }
{ "model": "deepseek/DeepSeek-V3.2", "finish_reason": "tool_calls", "status": null, "latency_ms": 1314, "time_to_first_token_ms": 398, "prompt_tokens": 2336, "completion_tokens": 46, "cost_usd": 0.00252, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 16 }
en
assistant_semantic:2:22:3
assistant_semantic
Include a check-in requirement involving timing, identification, payment card, or document. Exclude a general welcome message from a hotel.
absent
a301a100842580a4c9db7262
nemotron_interactive:5a1da92b553c3f03:10
core
nemotron_interactive
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_interactive:5a1da92b553c3f03:10
[ { "role": "system", "content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_de3a157fd5f2484b8470cad8", "name": "get_user_profile", "arguments": "{\"user_id\": \"789012\"}" } ], "tool_call_id": null, "name": null }
{ "model": "deepseek/DeepSeek-V3.2", "finish_reason": "tool_calls", "status": null, "latency_ms": 1314, "time_to_first_token_ms": 398, "prompt_tokens": 2336, "completion_tokens": 46, "cost_usd": 0.00252, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 16 }
en
quantitative_metadata:2:25:5
quantitative_metadata
The coding agent modified at least ten distinct files.
absent
df93a5723b9a1e61fd6c8f41
nemotron_interactive:5a1da92b553c3f03:10
core
nemotron_interactive
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_interactive:5a1da92b553c3f03:10
[ { "role": "system", "content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_de3a157fd5f2484b8470cad8", "name": "get_user_profile", "arguments": "{\"user_id\": \"789012\"}" } ], "tool_call_id": null, "name": null }
{ "model": "deepseek/DeepSeek-V3.2", "finish_reason": "tool_calls", "status": null, "latency_ms": 1314, "time_to_first_token_ms": 398, "prompt_tokens": 2336, "completion_tokens": 46, "cost_usd": 0.00252, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 16 }
en
assistant_semantic:3:10:0
assistant_semantic
offers human support
absent
20decba897f17067f9e887c0
nemotron_interactive:5a1da92b553c3f03:10
core
nemotron_interactive
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_interactive:5a1da92b553c3f03:10
[ { "role": "system", "content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_de3a157fd5f2484b8470cad8", "name": "get_user_profile", "arguments": "{\"user_id\": \"789012\"}" } ], "tool_call_id": null, "name": null }
{ "model": "deepseek/DeepSeek-V3.2", "finish_reason": "tool_calls", "status": null, "latency_ms": 1314, "time_to_first_token_ms": 398, "prompt_tokens": 2336, "completion_tokens": 46, "cost_usd": 0.00252, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 16 }
en
user_semantic:0:15:3
user_semantic
Include missing wages, incorrect deductions, PTO balance errors, and enrollment problems. Include requests to correct an identified compensation issue. Exclude a general question about benefit eligibility.
absent
d19c6f218c8fc84fda6d03cf
nemotron_interactive:5a1da92b553c3f03:10
core
nemotron_interactive
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_interactive:5a1da92b553c3f03:10
[ { "role": "system", "content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_de3a157fd5f2484b8470cad8", "name": "get_user_profile", "arguments": "{\"user_id\": \"789012\"}" } ], "tool_call_id": null, "name": null }
{ "model": "deepseek/DeepSeek-V3.2", "finish_reason": "tool_calls", "status": null, "latency_ms": 1314, "time_to_first_token_ms": 398, "prompt_tokens": 2336, "completion_tokens": 46, "cost_usd": 0.00252, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 16 }
en
quantitative_metadata:1:2:0
quantitative_metadata
too many available tools
present
96573f13bd56443ed5f08846
nemotron_interactive:5a1da92b553c3f03:10
core
nemotron_interactive
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_interactive:5a1da92b553c3f03:10
[ { "role": "system", "content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_de3a157fd5f2484b8470cad8", "name": "get_user_profile", "arguments": "{\"user_id\": \"789012\"}" } ], "tool_call_id": null, "name": null }
{ "model": "deepseek/DeepSeek-V3.2", "finish_reason": "tool_calls", "status": null, "latency_ms": 1314, "time_to_first_token_ms": 398, "prompt_tokens": 2336, "completion_tokens": 46, "cost_usd": 0.00252, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 16 }
en
user_semantic:2:15:3
user_semantic
Include stock or availability questions. Exclude requests for current travel availability.
present
3f714a2b8a702e24d6291c66
nemotron_interactive:5a1da92b553c3f03:10
core
nemotron_interactive
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_interactive:5a1da92b553c3f03:10
[ { "role": "system", "content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_de3a157fd5f2484b8470cad8", "name": "get_user_profile", "arguments": "{\"user_id\": \"789012\"}" } ], "tool_call_id": null, "name": null }
{ "model": "deepseek/DeepSeek-V3.2", "finish_reason": "tool_calls", "status": null, "latency_ms": 1314, "time_to_first_token_ms": 398, "prompt_tokens": 2336, "completion_tokens": 46, "cost_usd": 0.00252, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 16 }
en
user_semantic:2:15:2
user_semantic
Is the user asking whether an item is available?
present
4a0649ad732956d1ff77b425
nemotron_interactive:5a1da92b553c3f03:10
core
nemotron_interactive
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_interactive:5a1da92b553c3f03:10
[ { "role": "system", "content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_de3a157fd5f2484b8470cad8", "name": "get_user_profile", "arguments": "{\"user_id\": \"789012\"}" } ], "tool_call_id": null, "name": null }
{ "model": "deepseek/DeepSeek-V3.2", "finish_reason": "tool_calls", "status": null, "latency_ms": 1314, "time_to_first_token_ms": 398, "prompt_tokens": 2336, "completion_tokens": 46, "cost_usd": 0.00252, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 16 }
en
user_semantic:1:18:0
user_semantic
travel availability search
absent
ada16141b8b4b6ff428ed493
nemotron_interactive:5a1da92b553c3f03:10
core
nemotron_interactive
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_interactive:5a1da92b553c3f03:10
[ { "role": "system", "content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_de3a157fd5f2484b8470cad8", "name": "get_user_profile", "arguments": "{\"user_id\": \"789012\"}" } ], "tool_call_id": null, "name": null }
{ "model": "deepseek/DeepSeek-V3.2", "finish_reason": "tool_calls", "status": null, "latency_ms": 1314, "time_to_first_token_ms": 398, "prompt_tokens": 2336, "completion_tokens": 46, "cost_usd": 0.00252, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 16 }
en
assistant_semantic:1:16:5
assistant_semantic
multi-day itinerary
absent
2b1da77d17254fb3fe51005d
nemotron_interactive:5a1da92b553c3f03:10
core
nemotron_interactive
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_interactive:5a1da92b553c3f03:10
[ { "role": "system", "content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_de3a157fd5f2484b8470cad8", "name": "get_user_profile", "arguments": "{\"user_id\": \"789012\"}" } ], "tool_call_id": null, "name": null }
{ "model": "deepseek/DeepSeek-V3.2", "finish_reason": "tool_calls", "status": null, "latency_ms": 1314, "time_to_first_token_ms": 398, "prompt_tokens": 2336, "completion_tokens": 46, "cost_usd": 0.00252, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 16 }
en
user_semantic:0:10:2
user_semantic
Does the user include a phone number?
absent
95389c8cfae123f461a2f56b
nemotron_interactive:5a1da92b553c3f03:10
core
nemotron_interactive
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_interactive:5a1da92b553c3f03:10
[ { "role": "system", "content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "call_de3a157fd5f2484b8470cad8", "name": "get_user_profile", "arguments": "{\"user_id\": \"789012\"}" } ], "tool_call_id": null, "name": null }
{ "model": "deepseek/DeepSeek-V3.2", "finish_reason": "tool_calls", "status": null, "latency_ms": 1314, "time_to_first_token_ms": 398, "prompt_tokens": 2336, "completion_tokens": 46, "cost_usd": 0.00252, "temperature": null, "top_p": null, "seed": null, "stream": null, "tools_defined": 16 }
en
quantitative_metadata:0:12:0
quantitative_metadata
expensive LLM call
absent
dab603c9262268aac2e32613
nemotron_toolcall:11dacb6c428f1603:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:11dacb6c428f1603:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "Hey there. I’m curious about sous-vide pork shoulder for my taco pop-up. What settings do I need to get that perfect tender pull, and what safety considerations ar...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-bb9b038ebb5f35bf", "name": "youtube_search", "arguments": "{\"query\": \"sous vide pork shoulder tacos\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": "tool_calls", "status": "success", "latency_ms": 1566, "time_to_first_token_ms": 530, "prompt_tokens": null, "completion_tokens": 52, "cost_usd": 0.000703, "temperature": 0.5, "top_p": 1, "seed": 569482, "stream": null, "tools_defined": 7 }
en
surface_style:0:13:3
surface_style
Include replies of 10 or fewer visible words. Ignore citations, links, and tool metadata. Exclude a short reply followed by a substantive attachment.
present
42c93c3dfd1c2d32cacd5ab1
nemotron_toolcall:11dacb6c428f1603:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:11dacb6c428f1603:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "Hey there. I’m curious about sous-vide pork shoulder for my taco pop-up. What settings do I need to get that perfect tender pull, and what safety considerations ar...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-bb9b038ebb5f35bf", "name": "youtube_search", "arguments": "{\"query\": \"sous vide pork shoulder tacos\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": "tool_calls", "status": "success", "latency_ms": 1566, "time_to_first_token_ms": 530, "prompt_tokens": null, "completion_tokens": 52, "cost_usd": 0.000703, "temperature": 0.5, "top_p": 1, "seed": 569482, "stream": null, "tools_defined": 7 }
en
user_semantic:1:6:4
user_semantic
need the latest on this
absent
fd83cac83d6b2c1840b43c90
nemotron_toolcall:11dacb6c428f1603:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:11dacb6c428f1603:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "Hey there. I’m curious about sous-vide pork shoulder for my taco pop-up. What settings do I need to get that perfect tender pull, and what safety considerations ar...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-bb9b038ebb5f35bf", "name": "youtube_search", "arguments": "{\"query\": \"sous vide pork shoulder tacos\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": "tool_calls", "status": "success", "latency_ms": 1566, "time_to_first_token_ms": 530, "prompt_tokens": null, "completion_tokens": 52, "cost_usd": 0.000703, "temperature": 0.5, "top_p": 1, "seed": 569482, "stream": null, "tools_defined": 7 }
en
quantitative_metadata:0:12:2
quantitative_metadata
Did this call cost more than $1.00?
absent
a074a378fe828ca8001171b9
nemotron_toolcall:11dacb6c428f1603:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:11dacb6c428f1603:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "Hey there. I’m curious about sous-vide pork shoulder for my taco pop-up. What settings do I need to get that perfect tender pull, and what safety considerations ar...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-bb9b038ebb5f35bf", "name": "youtube_search", "arguments": "{\"query\": \"sous vide pork shoulder tacos\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": "tool_calls", "status": "success", "latency_ms": 1566, "time_to_first_token_ms": 530, "prompt_tokens": null, "completion_tokens": 52, "cost_usd": 0.000703, "temperature": 0.5, "top_p": 1, "seed": 569482, "stream": null, "tools_defined": 7 }
en
quantitative_metadata:2:20:2
quantitative_metadata
Did the assistant produce malformed arguments for a tool call?
absent
d10ebaec41153a5f679ccbc6
nemotron_toolcall:11dacb6c428f1603:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:11dacb6c428f1603:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "Hey there. I’m curious about sous-vide pork shoulder for my taco pop-up. What settings do I need to get that perfect tender pull, and what safety considerations ar...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-bb9b038ebb5f35bf", "name": "youtube_search", "arguments": "{\"query\": \"sous vide pork shoulder tacos\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": "tool_calls", "status": "success", "latency_ms": 1566, "time_to_first_token_ms": 530, "prompt_tokens": null, "completion_tokens": 52, "cost_usd": 0.000703, "temperature": 0.5, "top_p": 1, "seed": 569482, "stream": null, "tools_defined": 7 }
en
agentic_process:2:12:0
agentic_process
secret in tool input
absent
bba7d1dbe14d83db9a2c1a54
nemotron_toolcall:11dacb6c428f1603:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:11dacb6c428f1603:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "Hey there. I’m curious about sous-vide pork shoulder for my taco pop-up. What settings do I need to get that perfect tender pull, and what safety considerations ar...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-bb9b038ebb5f35bf", "name": "youtube_search", "arguments": "{\"query\": \"sous vide pork shoulder tacos\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": "tool_calls", "status": "success", "latency_ms": 1566, "time_to_first_token_ms": 530, "prompt_tokens": null, "completion_tokens": 52, "cost_usd": 0.000703, "temperature": 0.5, "top_p": 1, "seed": 569482, "stream": null, "tools_defined": 7 }
en
assistant_semantic:3:3:2
assistant_semantic
Is the assistant's answer too brief to be useful?
absent
630e15a765236d4a4c321210
nemotron_toolcall:11dacb6c428f1603:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:11dacb6c428f1603:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "Hey there. I’m curious about sous-vide pork shoulder for my taco pop-up. What settings do I need to get that perfect tender pull, and what safety considerations ar...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-bb9b038ebb5f35bf", "name": "youtube_search", "arguments": "{\"query\": \"sous vide pork shoulder tacos\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": "tool_calls", "status": "success", "latency_ms": 1566, "time_to_first_token_ms": 530, "prompt_tokens": null, "completion_tokens": 52, "cost_usd": 0.000703, "temperature": 0.5, "top_p": 1, "seed": 569482, "stream": null, "tools_defined": 7 }
en
user_semantic:0:19:1
user_semantic
The user says they cannot afford a payment, are struggling financially, or face serious money hardship.
absent
c59b7aa7e1c5d5de8f8c4ac9
nemotron_toolcall:11dacb6c428f1603:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:11dacb6c428f1603:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "Hey there. I’m curious about sous-vide pork shoulder for my taco pop-up. What settings do I need to get that perfect tender pull, and what safety considerations ar...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-bb9b038ebb5f35bf", "name": "youtube_search", "arguments": "{\"query\": \"sous vide pork shoulder tacos\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": "tool_calls", "status": "success", "latency_ms": 1566, "time_to_first_token_ms": 530, "prompt_tokens": null, "completion_tokens": 52, "cost_usd": 0.000703, "temperature": 0.5, "top_p": 1, "seed": 569482, "stream": null, "tools_defined": 7 }
en
user_semantic:2:15:4
user_semantic
do you guys have this in stock
absent
9c482044d6ccb05b2363abdb
nemotron_toolcall:11dacb6c428f1603:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:11dacb6c428f1603:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "Hey there. I’m curious about sous-vide pork shoulder for my taco pop-up. What settings do I need to get that perfect tender pull, and what safety considerations ar...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-bb9b038ebb5f35bf", "name": "youtube_search", "arguments": "{\"query\": \"sous vide pork shoulder tacos\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": "tool_calls", "status": "success", "latency_ms": 1566, "time_to_first_token_ms": 530, "prompt_tokens": null, "completion_tokens": 52, "cost_usd": 0.000703, "temperature": 0.5, "top_p": 1, "seed": 569482, "stream": null, "tools_defined": 7 }
en
relational:3:24:3
relational
Include when the user's latest message conveys frustration and refers to prior information being ignored or repeated. Exclude neutral restatements of information without dissatisfaction.
absent
cad9e6743f9c4f469320ffed
nemotron_toolcall:11dacb6c428f1603:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:11dacb6c428f1603:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "Hey there. I’m curious about sous-vide pork shoulder for my taco pop-up. What settings do I need to get that perfect tender pull, and what safety considerations ar...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-bb9b038ebb5f35bf", "name": "youtube_search", "arguments": "{\"query\": \"sous vide pork shoulder tacos\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": "tool_calls", "status": "success", "latency_ms": 1566, "time_to_first_token_ms": 530, "prompt_tokens": null, "completion_tokens": 52, "cost_usd": 0.000703, "temperature": 0.5, "top_p": 1, "seed": 569482, "stream": null, "tools_defined": 7 }
en
quantitative_metadata:1:2:0
quantitative_metadata
too many available tools
absent
dd9644e4bb70932da482bd24
nemotron_toolcall:11dacb6c428f1603:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:11dacb6c428f1603:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "Hey there. I’m curious about sous-vide pork shoulder for my taco pop-up. What settings do I need to get that perfect tender pull, and what safety considerations ar...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-bb9b038ebb5f35bf", "name": "youtube_search", "arguments": "{\"query\": \"sous vide pork shoulder tacos\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": "tool_calls", "status": "success", "latency_ms": 1566, "time_to_first_token_ms": 530, "prompt_tokens": null, "completion_tokens": 52, "cost_usd": 0.000703, "temperature": 0.5, "top_p": 1, "seed": 569482, "stream": null, "tools_defined": 7 }
en
agentic_process:2:8:4
agentic_process
tools disagreed and it just picked one
absent
50dc6f78da6d2b1699539f9b
nemotron_toolcall:11dacb6c428f1603:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:11dacb6c428f1603:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "Hey there. I’m curious about sous-vide pork shoulder for my taco pop-up. What settings do I need to get that perfect tender pull, and what safety considerations ar...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-bb9b038ebb5f35bf", "name": "youtube_search", "arguments": "{\"query\": \"sous vide pork shoulder tacos\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": "tool_calls", "status": "success", "latency_ms": 1566, "time_to_first_token_ms": 530, "prompt_tokens": null, "completion_tokens": 52, "cost_usd": 0.000703, "temperature": 0.5, "top_p": 1, "seed": 569482, "stream": null, "tools_defined": 7 }
en
quantitative_metadata:3:23:3
quantitative_metadata
Include when request metadata explicitly sets streaming to false. Exclude spans where streaming configuration is absent or unknown. Exclude streamed calls even if the final output was delivered quickly.
absent
e7d9f1896edecdbefbb7a487
nemotron_toolcall:11dacb6c428f1603:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:11dacb6c428f1603:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "Hey there. I’m curious about sous-vide pork shoulder for my taco pop-up. What settings do I need to get that perfect tender pull, and what safety considerations ar...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-bb9b038ebb5f35bf", "name": "youtube_search", "arguments": "{\"query\": \"sous vide pork shoulder tacos\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": "tool_calls", "status": "success", "latency_ms": 1566, "time_to_first_token_ms": 530, "prompt_tokens": null, "completion_tokens": 52, "cost_usd": 0.000703, "temperature": 0.5, "top_p": 1, "seed": 569482, "stream": null, "tools_defined": 7 }
en
agentic_process:2:4:0
agentic_process
irrelevant tool retrieval
absent
06a15a9b5917903dd6ff5644
nemotron_toolcall:11dacb6c428f1603:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:11dacb6c428f1603:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "Hey there. I’m curious about sous-vide pork shoulder for my taco pop-up. What settings do I need to get that perfect tender pull, and what safety considerations ar...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-bb9b038ebb5f35bf", "name": "youtube_search", "arguments": "{\"query\": \"sous vide pork shoulder tacos\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": "tool_calls", "status": "success", "latency_ms": 1566, "time_to_first_token_ms": 530, "prompt_tokens": null, "completion_tokens": 52, "cost_usd": 0.000703, "temperature": 0.5, "top_p": 1, "seed": 569482, "stream": null, "tools_defined": 7 }
en
surface_style:3:9:0
surface_style
user markdown table
absent
110e644db3797d348a180482
nemotron_toolcall:11dacb6c428f1603:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:11dacb6c428f1603:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "Hey there. I’m curious about sous-vide pork shoulder for my taco pop-up. What settings do I need to get that perfect tender pull, and what safety considerations ar...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-bb9b038ebb5f35bf", "name": "youtube_search", "arguments": "{\"query\": \"sous vide pork shoulder tacos\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": "tool_calls", "status": "success", "latency_ms": 1566, "time_to_first_token_ms": 530, "prompt_tokens": null, "completion_tokens": 52, "cost_usd": 0.000703, "temperature": 0.5, "top_p": 1, "seed": 569482, "stream": null, "tools_defined": 7 }
en
relational:3:7:1
relational
The assistant asks the user to choose again after the user had already selected an option and did not revisit that choice.
absent
22077a99fa7300365f245faa
nemotron_toolcall:11dacb6c428f1603:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:11dacb6c428f1603:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "Hey there. I’m curious about sous-vide pork shoulder for my taco pop-up. What settings do I need to get that perfect tender pull, and what safety considerations ar...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-bb9b038ebb5f35bf", "name": "youtube_search", "arguments": "{\"query\": \"sous vide pork shoulder tacos\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": "tool_calls", "status": "success", "latency_ms": 1566, "time_to_first_token_ms": 530, "prompt_tokens": null, "completion_tokens": 52, "cost_usd": 0.000703, "temperature": 0.5, "top_p": 1, "seed": 569482, "stream": null, "tools_defined": 7 }
en
assistant_semantic:0:6:4
assistant_semantic
pushes a particular option
absent
33781637f4c3b9b4c768058d
nemotron_toolcall:11dacb6c428f1603:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:11dacb6c428f1603:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "Hey there. I’m curious about sous-vide pork shoulder for my taco pop-up. What settings do I need to get that perfect tender pull, and what safety considerations ar...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-bb9b038ebb5f35bf", "name": "youtube_search", "arguments": "{\"query\": \"sous vide pork shoulder tacos\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": "tool_calls", "status": "success", "latency_ms": 1566, "time_to_first_token_ms": 530, "prompt_tokens": null, "completion_tokens": 52, "cost_usd": 0.000703, "temperature": 0.5, "top_p": 1, "seed": 569482, "stream": null, "tools_defined": 7 }
en
quantitative_metadata:1:2:4
quantitative_metadata
the model had a ridiculous menu of tools here
absent
ff35c8699172fb7225eb1b83
nemotron_toolcall:11dacb6c428f1603:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:11dacb6c428f1603:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "Hey there. I’m curious about sous-vide pork shoulder for my taco pop-up. What settings do I need to get that perfect tender pull, and what safety considerations ar...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-bb9b038ebb5f35bf", "name": "youtube_search", "arguments": "{\"query\": \"sous vide pork shoulder tacos\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": "tool_calls", "status": "success", "latency_ms": 1566, "time_to_first_token_ms": 530, "prompt_tokens": null, "completion_tokens": 52, "cost_usd": 0.000703, "temperature": 0.5, "top_p": 1, "seed": 569482, "stream": null, "tools_defined": 7 }
en
user_semantic:0:15:3
user_semantic
Include missing wages, incorrect deductions, PTO balance errors, and enrollment problems. Include requests to correct an identified compensation issue. Exclude a general question about benefit eligibility.
absent
abcf481ff545ce62793b2131
swe_smith:43a2d3852df03092:20
core
swe_smith
Kwai-Klear/SWE-smith-mini_swe_agent_plus-trajectories-66k
mit
swe_smith:43a2d3852df03092:20
[ { "role": "system", "content": "You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.\nYour response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).\n\nInclude a THOUGHT section before your command wher...
{ "role": "assistant", "content": "THOUGHT: Let me continue reading the output.py file to see more of the Output class definition and understand how it handles version-aware functionality.", "tool_calls": [ { "id": "call_10", "name": "bash", "arguments": "{\"command\": \"sed -n '100,200p' dv...
{ "model": "swe-smith-agent", "finish_reason": "tool_calls", "status": "success", "latency_ms": 3545, "time_to_first_token_ms": 563, "prompt_tokens": 10113, "completion_tokens": null, "cost_usd": 0.010461, "temperature": 1, "top_p": 0.95, "seed": null, "stream": null, "tools_defined": 1 }
en
quantitative_metadata:2:25:2
quantitative_metadata
Did the agent change three or more files?
absent
9eceb8e9920b159fe4f9a04c
swe_smith:43a2d3852df03092:20
core
swe_smith
Kwai-Klear/SWE-smith-mini_swe_agent_plus-trajectories-66k
mit
swe_smith:43a2d3852df03092:20
[ { "role": "system", "content": "You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.\nYour response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).\n\nInclude a THOUGHT section before your command wher...
{ "role": "assistant", "content": "THOUGHT: Let me continue reading the output.py file to see more of the Output class definition and understand how it handles version-aware functionality.", "tool_calls": [ { "id": "call_10", "name": "bash", "arguments": "{\"command\": \"sed -n '100,200p' dv...
{ "model": "swe-smith-agent", "finish_reason": "tool_calls", "status": "success", "latency_ms": 3545, "time_to_first_token_ms": 563, "prompt_tokens": 10113, "completion_tokens": null, "cost_usd": 0.010461, "temperature": 1, "top_p": 0.95, "seed": null, "stream": null, "tools_defined": 1 }
en
relational:2:17:4
relational
they switched languages and the bot just keeps answering in the old one
absent
80128ef059c097ef2640ecbc
swe_smith:43a2d3852df03092:20
core
swe_smith
Kwai-Klear/SWE-smith-mini_swe_agent_plus-trajectories-66k
mit
swe_smith:43a2d3852df03092:20
[ { "role": "system", "content": "You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.\nYour response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).\n\nInclude a THOUGHT section before your command wher...
{ "role": "assistant", "content": "THOUGHT: Let me continue reading the output.py file to see more of the Output class definition and understand how it handles version-aware functionality.", "tool_calls": [ { "id": "call_10", "name": "bash", "arguments": "{\"command\": \"sed -n '100,200p' dv...
{ "model": "swe-smith-agent", "finish_reason": "tool_calls", "status": "success", "latency_ms": 3545, "time_to_first_token_ms": 563, "prompt_tokens": 10113, "completion_tokens": null, "cost_usd": 0.010461, "temperature": 1, "top_p": 0.95, "seed": null, "stream": null, "tools_defined": 1 }
en
relational:3:6:2
relational
Did the tutor continue using terminology the learner had just corrected?
absent
eac72e1717f335684f97a720
swe_smith:43a2d3852df03092:20
core
swe_smith
Kwai-Klear/SWE-smith-mini_swe_agent_plus-trajectories-66k
mit
swe_smith:43a2d3852df03092:20
[ { "role": "system", "content": "You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.\nYour response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).\n\nInclude a THOUGHT section before your command wher...
{ "role": "assistant", "content": "THOUGHT: Let me continue reading the output.py file to see more of the Output class definition and understand how it handles version-aware functionality.", "tool_calls": [ { "id": "call_10", "name": "bash", "arguments": "{\"command\": \"sed -n '100,200p' dv...
{ "model": "swe-smith-agent", "finish_reason": "tool_calls", "status": "success", "latency_ms": 3545, "time_to_first_token_ms": 563, "prompt_tokens": 10113, "completion_tokens": null, "cost_usd": 0.010461, "temperature": 1, "top_p": 0.95, "seed": null, "stream": null, "tools_defined": 1 }
en
assistant_semantic:2:7:0
assistant_semantic
credits user input
absent
7dd8e1a70ea0b53328dae68c
swe_smith:43a2d3852df03092:20
core
swe_smith
Kwai-Klear/SWE-smith-mini_swe_agent_plus-trajectories-66k
mit
swe_smith:43a2d3852df03092:20
[ { "role": "system", "content": "You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.\nYour response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).\n\nInclude a THOUGHT section before your command wher...
{ "role": "assistant", "content": "THOUGHT: Let me continue reading the output.py file to see more of the Output class definition and understand how it handles version-aware functionality.", "tool_calls": [ { "id": "call_10", "name": "bash", "arguments": "{\"command\": \"sed -n '100,200p' dv...
{ "model": "swe-smith-agent", "finish_reason": "tool_calls", "status": "success", "latency_ms": 3545, "time_to_first_token_ms": 563, "prompt_tokens": 10113, "completion_tokens": null, "cost_usd": 0.010461, "temperature": 1, "top_p": 0.95, "seed": null, "stream": null, "tools_defined": 1 }
en
assistant_semantic:0:4:4
assistant_semantic
says it already did the thing
absent
ad6e340199dd009b9922f795
swe_smith:43a2d3852df03092:20
core
swe_smith
Kwai-Klear/SWE-smith-mini_swe_agent_plus-trajectories-66k
mit
swe_smith:43a2d3852df03092:20
[ { "role": "system", "content": "You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.\nYour response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).\n\nInclude a THOUGHT section before your command wher...
{ "role": "assistant", "content": "THOUGHT: Let me continue reading the output.py file to see more of the Output class definition and understand how it handles version-aware functionality.", "tool_calls": [ { "id": "call_10", "name": "bash", "arguments": "{\"command\": \"sed -n '100,200p' dv...
{ "model": "swe-smith-agent", "finish_reason": "tool_calls", "status": "success", "latency_ms": 3545, "time_to_first_token_ms": 563, "prompt_tokens": 10113, "completion_tokens": null, "cost_usd": 0.010461, "temperature": 1, "top_p": 0.95, "seed": null, "stream": null, "tools_defined": 1 }
en
agentic_process:2:4:2
agentic_process
Did the agent use a tool to fetch unrelated information rather than information needed for the task?
absent
c4368012c244c61b69c83513
swe_smith:43a2d3852df03092:20
core
swe_smith
Kwai-Klear/SWE-smith-mini_swe_agent_plus-trajectories-66k
mit
swe_smith:43a2d3852df03092:20
[ { "role": "system", "content": "You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.\nYour response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).\n\nInclude a THOUGHT section before your command wher...
{ "role": "assistant", "content": "THOUGHT: Let me continue reading the output.py file to see more of the Output class definition and understand how it handles version-aware functionality.", "tool_calls": [ { "id": "call_10", "name": "bash", "arguments": "{\"command\": \"sed -n '100,200p' dv...
{ "model": "swe-smith-agent", "finish_reason": "tool_calls", "status": "success", "latency_ms": 3545, "time_to_first_token_ms": 563, "prompt_tokens": 10113, "completion_tokens": null, "cost_usd": 0.010461, "temperature": 1, "top_p": 0.95, "seed": null, "stream": null, "tools_defined": 1 }
en
relational:1:5:2
relational
Does the assistant use a previously defined term with a different meaning?
absent
41f3c50e2add6cf4c535002f
swe_smith:43a2d3852df03092:20
core
swe_smith
Kwai-Klear/SWE-smith-mini_swe_agent_plus-trajectories-66k
mit
swe_smith:43a2d3852df03092:20
[ { "role": "system", "content": "You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.\nYour response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).\n\nInclude a THOUGHT section before your command wher...
{ "role": "assistant", "content": "THOUGHT: Let me continue reading the output.py file to see more of the Output class definition and understand how it handles version-aware functionality.", "tool_calls": [ { "id": "call_10", "name": "bash", "arguments": "{\"command\": \"sed -n '100,200p' dv...
{ "model": "swe-smith-agent", "finish_reason": "tool_calls", "status": "success", "latency_ms": 3545, "time_to_first_token_ms": 563, "prompt_tokens": 10113, "completion_tokens": null, "cost_usd": 0.010461, "temperature": 1, "top_p": 0.95, "seed": null, "stream": null, "tools_defined": 1 }
en
surface_style:0:19:3
surface_style
Include text enclosed in single backticks. Include inline variables or commands. Exclude triple-backtick fenced blocks.
absent
b57e1226ad0526f3fb309f51
swe_smith:43a2d3852df03092:20
core
swe_smith
Kwai-Klear/SWE-smith-mini_swe_agent_plus-trajectories-66k
mit
swe_smith:43a2d3852df03092:20
[ { "role": "system", "content": "You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.\nYour response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).\n\nInclude a THOUGHT section before your command wher...
{ "role": "assistant", "content": "THOUGHT: Let me continue reading the output.py file to see more of the Output class definition and understand how it handles version-aware functionality.", "tool_calls": [ { "id": "call_10", "name": "bash", "arguments": "{\"command\": \"sed -n '100,200p' dv...
{ "model": "swe-smith-agent", "finish_reason": "tool_calls", "status": "success", "latency_ms": 3545, "time_to_first_token_ms": 563, "prompt_tokens": 10113, "completion_tokens": null, "cost_usd": 0.010461, "temperature": 1, "top_p": 0.95, "seed": null, "stream": null, "tools_defined": 1 }
en
agentic_process:2:12:5
agentic_process
Flag only secrets passed to a third-party or external tool.
absent
5f26f76142e714b1ab2dfd68
swe_smith:43a2d3852df03092:20
core
swe_smith
Kwai-Klear/SWE-smith-mini_swe_agent_plus-trajectories-66k
mit
swe_smith:43a2d3852df03092:20
[ { "role": "system", "content": "You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.\nYour response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).\n\nInclude a THOUGHT section before your command wher...
{ "role": "assistant", "content": "THOUGHT: Let me continue reading the output.py file to see more of the Output class definition and understand how it handles version-aware functionality.", "tool_calls": [ { "id": "call_10", "name": "bash", "arguments": "{\"command\": \"sed -n '100,200p' dv...
{ "model": "swe-smith-agent", "finish_reason": "tool_calls", "status": "success", "latency_ms": 3545, "time_to_first_token_ms": 563, "prompt_tokens": 10113, "completion_tokens": null, "cost_usd": 0.010461, "temperature": 1, "top_p": 0.95, "seed": null, "stream": null, "tools_defined": 1 }
en
agentic_process:2:12:0
agentic_process
secret in tool input
absent
03af8400284dc999d87c0026
swe_smith:43a2d3852df03092:20
core
swe_smith
Kwai-Klear/SWE-smith-mini_swe_agent_plus-trajectories-66k
mit
swe_smith:43a2d3852df03092:20
[ { "role": "system", "content": "You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.\nYour response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).\n\nInclude a THOUGHT section before your command wher...
{ "role": "assistant", "content": "THOUGHT: Let me continue reading the output.py file to see more of the Output class definition and understand how it handles version-aware functionality.", "tool_calls": [ { "id": "call_10", "name": "bash", "arguments": "{\"command\": \"sed -n '100,200p' dv...
{ "model": "swe-smith-agent", "finish_reason": "tool_calls", "status": "success", "latency_ms": 3545, "time_to_first_token_ms": 563, "prompt_tokens": 10113, "completion_tokens": null, "cost_usd": 0.010461, "temperature": 1, "top_p": 0.95, "seed": null, "stream": null, "tools_defined": 1 }
en
quantitative_metadata:2:8:4
quantitative_metadata
this response was run with a seed
not_observable
7461276e67ca37cfe594b32e
swe_smith:43a2d3852df03092:20
core
swe_smith
Kwai-Klear/SWE-smith-mini_swe_agent_plus-trajectories-66k
mit
swe_smith:43a2d3852df03092:20
[ { "role": "system", "content": "You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.\nYour response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).\n\nInclude a THOUGHT section before your command wher...
{ "role": "assistant", "content": "THOUGHT: Let me continue reading the output.py file to see more of the Output class definition and understand how it handles version-aware functionality.", "tool_calls": [ { "id": "call_10", "name": "bash", "arguments": "{\"command\": \"sed -n '100,200p' dv...
{ "model": "swe-smith-agent", "finish_reason": "tool_calls", "status": "success", "latency_ms": 3545, "time_to_first_token_ms": 563, "prompt_tokens": 10113, "completion_tokens": null, "cost_usd": 0.010461, "temperature": 1, "top_p": 0.95, "seed": null, "stream": null, "tools_defined": 1 }
en
assistant_semantic:3:6:3
assistant_semantic
Include when the output implies the user is selfish, irresponsible, or letting others down unless they act. Exclude ordinary explanations of consequences without emotional pressure.
absent
72da2696d10bfdcbb30f708f
swe_smith:43a2d3852df03092:20
core
swe_smith
Kwai-Klear/SWE-smith-mini_swe_agent_plus-trajectories-66k
mit
swe_smith:43a2d3852df03092:20
[ { "role": "system", "content": "You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.\nYour response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).\n\nInclude a THOUGHT section before your command wher...
{ "role": "assistant", "content": "THOUGHT: Let me continue reading the output.py file to see more of the Output class definition and understand how it handles version-aware functionality.", "tool_calls": [ { "id": "call_10", "name": "bash", "arguments": "{\"command\": \"sed -n '100,200p' dv...
{ "model": "swe-smith-agent", "finish_reason": "tool_calls", "status": "success", "latency_ms": 3545, "time_to_first_token_ms": 563, "prompt_tokens": 10113, "completion_tokens": null, "cost_usd": 0.010461, "temperature": 1, "top_p": 0.95, "seed": null, "stream": null, "tools_defined": 1 }
en
quantitative_metadata:2:25:0
quantitative_metadata
three files changed
absent
8d9f4fdd049b4c126ef188f5
swe_smith:43a2d3852df03092:20
core
swe_smith
Kwai-Klear/SWE-smith-mini_swe_agent_plus-trajectories-66k
mit
swe_smith:43a2d3852df03092:20
[ { "role": "system", "content": "You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.\nYour response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).\n\nInclude a THOUGHT section before your command wher...
{ "role": "assistant", "content": "THOUGHT: Let me continue reading the output.py file to see more of the Output class definition and understand how it handles version-aware functionality.", "tool_calls": [ { "id": "call_10", "name": "bash", "arguments": "{\"command\": \"sed -n '100,200p' dv...
{ "model": "swe-smith-agent", "finish_reason": "tool_calls", "status": "success", "latency_ms": 3545, "time_to_first_token_ms": 563, "prompt_tokens": 10113, "completion_tokens": null, "cost_usd": 0.010461, "temperature": 1, "top_p": 0.95, "seed": null, "stream": null, "tools_defined": 1 }
en
surface_style:2:21:2
surface_style
Does a single line occur three or more times in the latest tool result?
absent
6e3ccec8fae26b7a1fe07f7f
swe_smith:43a2d3852df03092:20
core
swe_smith
Kwai-Klear/SWE-smith-mini_swe_agent_plus-trajectories-66k
mit
swe_smith:43a2d3852df03092:20
[ { "role": "system", "content": "You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.\nYour response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).\n\nInclude a THOUGHT section before your command wher...
{ "role": "assistant", "content": "THOUGHT: Let me continue reading the output.py file to see more of the Output class definition and understand how it handles version-aware functionality.", "tool_calls": [ { "id": "call_10", "name": "bash", "arguments": "{\"command\": \"sed -n '100,200p' dv...
{ "model": "swe-smith-agent", "finish_reason": "tool_calls", "status": "success", "latency_ms": 3545, "time_to_first_token_ms": 563, "prompt_tokens": 10113, "completion_tokens": null, "cost_usd": 0.010461, "temperature": 1, "top_p": 0.95, "seed": null, "stream": null, "tools_defined": 1 }
en
assistant_semantic:0:17:3
assistant_semantic
Include: guidance on leave, benefits, conduct, pay, performance, or workplace rules. Exclude: directing the user to HR without explaining policy.
absent
98a5a60f791977eddccc77bf
swe_smith:43a2d3852df03092:20
core
swe_smith
Kwai-Klear/SWE-smith-mini_swe_agent_plus-trajectories-66k
mit
swe_smith:43a2d3852df03092:20
[ { "role": "system", "content": "You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.\nYour response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).\n\nInclude a THOUGHT section before your command wher...
{ "role": "assistant", "content": "THOUGHT: Let me continue reading the output.py file to see more of the Output class definition and understand how it handles version-aware functionality.", "tool_calls": [ { "id": "call_10", "name": "bash", "arguments": "{\"command\": \"sed -n '100,200p' dv...
{ "model": "swe-smith-agent", "finish_reason": "tool_calls", "status": "success", "latency_ms": 3545, "time_to_first_token_ms": 563, "prompt_tokens": 10113, "completion_tokens": null, "cost_usd": 0.010461, "temperature": 1, "top_p": 0.95, "seed": null, "stream": null, "tools_defined": 1 }
en
surface_style:2:21:0
surface_style
duplicate tool output lines
absent
2e23044a132ed65f9e8e200e
swe_smith:43a2d3852df03092:20
core
swe_smith
Kwai-Klear/SWE-smith-mini_swe_agent_plus-trajectories-66k
mit
swe_smith:43a2d3852df03092:20
[ { "role": "system", "content": "You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.\nYour response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).\n\nInclude a THOUGHT section before your command wher...
{ "role": "assistant", "content": "THOUGHT: Let me continue reading the output.py file to see more of the Output class definition and understand how it handles version-aware functionality.", "tool_calls": [ { "id": "call_10", "name": "bash", "arguments": "{\"command\": \"sed -n '100,200p' dv...
{ "model": "swe-smith-agent", "finish_reason": "tool_calls", "status": "success", "latency_ms": 3545, "time_to_first_token_ms": 563, "prompt_tokens": 10113, "completion_tokens": null, "cost_usd": 0.010461, "temperature": 1, "top_p": 0.95, "seed": null, "stream": null, "tools_defined": 1 }
en
agentic_process:1:6:5
agentic_process
Flag only direct interpolation of tool output into a shell command.
absent
62b4e655e04e9c076ae1c585
swe_smith:43a2d3852df03092:20
core
swe_smith
Kwai-Klear/SWE-smith-mini_swe_agent_plus-trajectories-66k
mit
swe_smith:43a2d3852df03092:20
[ { "role": "system", "content": "You are a helpful assistant that can interact multiple times with a computer shell to solve programming tasks.\nYour response must contain exactly ONE bash code block with ONE command (or commands connected with && or ||).\n\nInclude a THOUGHT section before your command wher...
{ "role": "assistant", "content": "THOUGHT: Let me continue reading the output.py file to see more of the Output class definition and understand how it handles version-aware functionality.", "tool_calls": [ { "id": "call_10", "name": "bash", "arguments": "{\"command\": \"sed -n '100,200p' dv...
{ "model": "swe-smith-agent", "finish_reason": "tool_calls", "status": "success", "latency_ms": 3545, "time_to_first_token_ms": 563, "prompt_tokens": 10113, "completion_tokens": null, "cost_usd": 0.010461, "temperature": 1, "top_p": 0.95, "seed": null, "stream": null, "tools_defined": 1 }
en
quantitative_metadata:1:2:0
quantitative_metadata
too many available tools
absent
ac0545c003f4f3e59cf16a50
nemotron_toolcall:c7dc8f1f077eed0e:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:c7dc8f1f077eed0e:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "I'm helping plan a heritage trip that includes the Darjeeling Himalayan Railway and the Nilgiri Mountain Railway, but I need to figure out the route and schedule f...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-8121480ec9a4cf9b", "name": "indian_railway_search", "arguments": "{\"search\": \"Nilgiri Mountain Railway\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": null, "status": "success", "latency_ms": 1124, "time_to_first_token_ms": 488, "prompt_tokens": 924, "completion_tokens": null, "cost_usd": 0.001136, "temperature": 0, "top_p": 0.9, "seed": 252109, "stream": null, "tools_defined": 1 }
en
surface_style:3:9:4
surface_style
they sent the trip options as a markdown table
absent
a29c13852c0424a7c11b8297
nemotron_toolcall:c7dc8f1f077eed0e:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:c7dc8f1f077eed0e:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "I'm helping plan a heritage trip that includes the Darjeeling Himalayan Railway and the Nilgiri Mountain Railway, but I need to figure out the route and schedule f...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-8121480ec9a4cf9b", "name": "indian_railway_search", "arguments": "{\"search\": \"Nilgiri Mountain Railway\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": null, "status": "success", "latency_ms": 1124, "time_to_first_token_ms": 488, "prompt_tokens": 924, "completion_tokens": null, "cost_usd": 0.001136, "temperature": 0, "top_p": 0.9, "seed": 252109, "stream": null, "tools_defined": 1 }
en
user_semantic:1:22:0
user_semantic
transaction status check
absent
a5a8c47a999e05b7f4619802
nemotron_toolcall:c7dc8f1f077eed0e:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:c7dc8f1f077eed0e:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "I'm helping plan a heritage trip that includes the Darjeeling Himalayan Railway and the Nilgiri Mountain Railway, but I need to figure out the route and schedule f...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-8121480ec9a4cf9b", "name": "indian_railway_search", "arguments": "{\"search\": \"Nilgiri Mountain Railway\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": null, "status": "success", "latency_ms": 1124, "time_to_first_token_ms": 488, "prompt_tokens": 924, "completion_tokens": null, "cost_usd": 0.001136, "temperature": 0, "top_p": 0.9, "seed": 252109, "stream": null, "tools_defined": 1 }
en
assistant_semantic:1:16:2
assistant_semantic
Does the assistant provide a travel itinerary?
absent
8b023417326667dffe1928e6
nemotron_toolcall:c7dc8f1f077eed0e:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:c7dc8f1f077eed0e:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "I'm helping plan a heritage trip that includes the Darjeeling Himalayan Railway and the Nilgiri Mountain Railway, but I need to figure out the route and schedule f...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-8121480ec9a4cf9b", "name": "indian_railway_search", "arguments": "{\"search\": \"Nilgiri Mountain Railway\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": null, "status": "success", "latency_ms": 1124, "time_to_first_token_ms": 488, "prompt_tokens": 924, "completion_tokens": null, "cost_usd": 0.001136, "temperature": 0, "top_p": 0.9, "seed": 252109, "stream": null, "tools_defined": 1 }
en
quantitative_metadata:2:3:3
quantitative_metadata
Include exact o3 model identifiers. Exclude o3-mini, o1, and general-purpose GPT models.
absent
a6b097cafbe393b15518a6c7
nemotron_toolcall:c7dc8f1f077eed0e:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:c7dc8f1f077eed0e:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "I'm helping plan a heritage trip that includes the Darjeeling Himalayan Railway and the Nilgiri Mountain Railway, but I need to figure out the route and schedule f...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-8121480ec9a4cf9b", "name": "indian_railway_search", "arguments": "{\"search\": \"Nilgiri Mountain Railway\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": null, "status": "success", "latency_ms": 1124, "time_to_first_token_ms": 488, "prompt_tokens": 924, "completion_tokens": null, "cost_usd": 0.001136, "temperature": 0, "top_p": 0.9, "seed": 252109, "stream": null, "tools_defined": 1 }
en
user_semantic:1:14:4
user_semantic
i was born on 04/12/1988
absent
8326cc0d3a577466c270d550
nemotron_toolcall:c7dc8f1f077eed0e:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:c7dc8f1f077eed0e:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "I'm helping plan a heritage trip that includes the Darjeeling Himalayan Railway and the Nilgiri Mountain Railway, but I need to figure out the route and schedule f...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-8121480ec9a4cf9b", "name": "indian_railway_search", "arguments": "{\"search\": \"Nilgiri Mountain Railway\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": null, "status": "success", "latency_ms": 1124, "time_to_first_token_ms": 488, "prompt_tokens": 924, "completion_tokens": null, "cost_usd": 0.001136, "temperature": 0, "top_p": 0.9, "seed": 252109, "stream": null, "tools_defined": 1 }
en
user_semantic:2:15:3
user_semantic
Include stock or availability questions. Exclude requests for current travel availability.
absent
7dafda2ca614a47ffc761639
nemotron_toolcall:c7dc8f1f077eed0e:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:c7dc8f1f077eed0e:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "I'm helping plan a heritage trip that includes the Darjeeling Himalayan Railway and the Nilgiri Mountain Railway, but I need to figure out the route and schedule f...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-8121480ec9a4cf9b", "name": "indian_railway_search", "arguments": "{\"search\": \"Nilgiri Mountain Railway\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": null, "status": "success", "latency_ms": 1124, "time_to_first_token_ms": 488, "prompt_tokens": 924, "completion_tokens": null, "cost_usd": 0.001136, "temperature": 0, "top_p": 0.9, "seed": 252109, "stream": null, "tools_defined": 1 }
en
assistant_semantic:2:22:1
assistant_semantic
The assistant states a requirement for hotel, flight, rail, or rental-car check-in.
absent
6510c148e2693d17025853d5
nemotron_toolcall:c7dc8f1f077eed0e:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:c7dc8f1f077eed0e:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "I'm helping plan a heritage trip that includes the Darjeeling Himalayan Railway and the Nilgiri Mountain Railway, but I need to figure out the route and schedule f...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-8121480ec9a4cf9b", "name": "indian_railway_search", "arguments": "{\"search\": \"Nilgiri Mountain Railway\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": null, "status": "success", "latency_ms": 1124, "time_to_first_token_ms": 488, "prompt_tokens": 924, "completion_tokens": null, "cost_usd": 0.001136, "temperature": 0, "top_p": 0.9, "seed": 252109, "stream": null, "tools_defined": 1 }
en
surface_style:2:21:2
surface_style
Does a single line occur three or more times in the latest tool result?
absent
29819ac98f71cef5e7b460ee
nemotron_toolcall:c7dc8f1f077eed0e:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:c7dc8f1f077eed0e:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "I'm helping plan a heritage trip that includes the Darjeeling Himalayan Railway and the Nilgiri Mountain Railway, but I need to figure out the route and schedule f...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-8121480ec9a4cf9b", "name": "indian_railway_search", "arguments": "{\"search\": \"Nilgiri Mountain Railway\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": null, "status": "success", "latency_ms": 1124, "time_to_first_token_ms": 488, "prompt_tokens": 924, "completion_tokens": null, "cost_usd": 0.001136, "temperature": 0, "top_p": 0.9, "seed": 252109, "stream": null, "tools_defined": 1 }
en
assistant_semantic:1:16:5
assistant_semantic
multi-day itinerary
absent
b79cb62a9df90f6bd7832337
nemotron_toolcall:c7dc8f1f077eed0e:4
core
nemotron_toolcall
nvidia/Nemotron-SFT-Agentic-v2
cc-by-4.0
nemotron_toolcall:c7dc8f1f077eed0e:4
[ { "role": "system", "content": "", "tool_calls": [], "tool_call_id": null, "name": null }, { "role": "user", "content": "I'm helping plan a heritage trip that includes the Darjeeling Himalayan Railway and the Nilgiri Mountain Railway, but I need to figure out the route and schedule f...
{ "role": "assistant", "content": "", "tool_calls": [ { "id": "chatcmpl-tool-8121480ec9a4cf9b", "name": "indian_railway_search", "arguments": "{\"search\": \"Nilgiri Mountain Railway\"}" } ], "tool_call_id": null, "name": null }
{ "model": "DeepSeek-v3.2", "finish_reason": null, "status": "success", "latency_ms": 1124, "time_to_first_token_ms": 488, "prompt_tokens": 924, "completion_tokens": null, "cost_usd": 0.001136, "temperature": 0, "top_p": 0.9, "seed": 252109, "stream": null, "tools_defined": 1 }
en
agentic_process:3:16:1
agentic_process
The helpdesk agent adds or removes a user from a group without first checking whether the user is already in the intended membership state.
absent