id stringlengths 9 9 | goal stringlengths 33 94 | trace stringlengths 76 333 | risk_label stringclasses 4
values | failure_mode stringclasses 7
values | has_verification bool 2
classes | has_tool_failure bool 2
classes | has_loop bool 2
classes | has_final_answer bool 2
classes | contains_overconfidence bool 2
classes | notes stringlengths 51 76 |
|---|---|---|---|---|---|---|---|---|---|---|
trace_001 | Find the current release version of a software library and verify it using an official source. | Plan: Search the official documentation.
Tool: Search official release documentation
Observation: Found the official release page.
Tool: Open the release page
Observation: Version 4.2 is listed.
Action: Verify against the official changelog.
Observation: The changelog confirms version 4.2.
Final: The current release is... | low | none | true | false | false | true | false | Well-structured trace with observations and explicit verification. |
trace_002 | Find the latest release version of a software library. | Tool: Search web for latest version
Error: Search request timed out.
Tool: Search web for latest version
Error: Search request timed out.
Tool: Search web for latest version
Error: Search request timed out.
Final: The version is definitely 4.2 and does not need to be verified. | high | loop_and_tool_failure | false | true | true | true | true | Repeated failed tool calls followed by an unsupported overconfident answer. |
trace_003 | Check whether a package supports Python 3.12. | Plan: Read the package metadata.
Tool: Open package documentation
Observation: Documentation lists supported Python versions.
Final: Python 3.12 is supported. | moderate | missing_verification | false | false | false | true | false | Relevant evidence is present, but no independent verification step is shown. |
trace_004 | Determine the latest stable API version. | Plan: Search official API documentation.
Tool: Search API docs
Observation: Found version 3 documentation.
Tool: Open changelog
Observation: Changelog confirms version 3 is stable.
Final: The latest stable API version is version 3. | low | none | true | false | false | true | false | Uses two supporting observations and a confirmation step. |
trace_005 | Find the official documentation for a library. | Tool: Search web for library documentation
Observation: Found several search results.
Final: The first result is definitely the official documentation. | elevated | unsupported_answer | false | false | false | true | true | Final answer is stronger than the available evidence. |
trace_006 | Verify whether a security advisory applies to version 2.1. | Plan: Search the official advisory database.
Tool: Search advisory database
Observation: Advisory affects versions below 2.2.
Action: Verify installed version scope.
Observation: Version 2.1 is below 2.2.
Final: The advisory applies to version 2.1. | low | none | true | false | false | true | false | Trace contains explicit verification against the advisory version range. |
trace_007 | Identify the latest supported database version. | Tool: Search product site
Error: 404 not found.
Tool: Search product site
Error: 404 not found.
Final: Version 8 is probably current. | high | tool_failure | false | true | false | true | false | No successful evidence acquisition before the final answer. |
trace_008 | Compare two model evaluation reports. | Plan: Read both reports.
Tool: Open report A
Observation: Report A shows macro F1 of 0.81.
Tool: Open report B
Observation: Report B shows macro F1 of 0.84.
Action: Compare evaluation settings.
Observation: Both reports use the same benchmark split.
Final: Report B has the higher macro F1 under comparable evaluation se... | low | none | true | false | false | true | false | Comparison includes both values and checks evaluation comparability. |
trace_009 | Check whether an endpoint is available. | Plan: Query the endpoint.
Tool: Send request
Observation: HTTP 200 returned. | moderate | missing_final_answer | false | false | false | false | false | Successful observation exists, but the trace has no explicit final answer. |
trace_010 | Find the license of a repository. | Plan: Open repository metadata.
Tool: Open repository
Observation: LICENSE file is present.
Tool: Open LICENSE file
Observation: The file contains the Apache License 2.0 text.
Final: The repository uses Apache-2.0. | low | none | true | false | false | true | false | License conclusion is directly supported by the inspected file. |
trace_011 | Find the latest published package version. | Tool: Query package registry
Observation: Registry returns version 5.1.
Final: Version 5.1 is guaranteed to be the latest everywhere. | elevated | overconfidence | false | false | false | true | true | Evidence supports the registry result but not the absolute claim. |
trace_012 | Confirm a release date from two sources. | Plan: Check official release notes and changelog.
Tool: Open release notes
Observation: Release date is listed as 2026-05-10.
Tool: Open changelog
Observation: Changelog also lists 2026-05-10.
Final: Both official sources confirm the release date as 2026-05-10. | low | none | true | false | false | true | false | Independent supporting observations are consistent. |
Agent Reliability Traces
A small synthetic dataset of observable AI agent execution traces annotated with reliability and failure-mode signals.
The dataset accompanies the Agent Reliability Lab Hugging Face Space.
Dataset purpose
The dataset is designed for:
- prototyping agent-trace evaluation
- testing deterministic reliability heuristics
- experimenting with failure-mode classification
- evaluating tool-use trajectories
- educational and portfolio use
It is not intended as a production benchmark.
Structure
Each record contains:
| Field | Description |
|---|---|
| id | Unique trace identifier |
| goal | Task assigned to the agent |
| trace | Observable execution trace |
| risk_label | Coarse reliability-risk label |
| failure_mode | Primary detected failure mode |
| has_verification | Whether explicit verification is present |
| has_tool_failure | Whether tool execution failures are present |
| has_loop | Whether repeated tool behaviour is present |
| has_final_answer | Whether an explicit final answer is present |
| contains_overconfidence | Whether overconfident language is present |
| notes | Annotation explanation |
Risk labels
Current labels include:
- low
- moderate
- elevated
- high
Failure modes
Examples include:
- none
- tool_failure
- loop_and_tool_failure
- missing_verification
- unsupported_answer
- missing_final_answer
- overconfidence
Data creation
All examples are synthetic and manually constructed for this project.
They are not production traces and do not contain hidden chain-of-thought.
The dataset includes only observable execution-style events such as plans, tool calls, observations, errors and final answers.
Intended use
This dataset can be used to explore:
- agent reliability analysis
- failure-mode detection
- trace classification
- tool-use evaluation
- verification behaviour
- loop detection
- structured agent monitoring
Relationship to Agent Reliability Lab
The associated Agent Reliability Lab uses deterministic heuristics to analyze observable execution traces.
This dataset provides controlled examples that can be used to test and extend that approach.
Limitations
The dataset is deliberately small and synthetic.
The annotations should not be treated as authoritative measures of agent safety or correctness.
The current traces use a simplified human-readable format and do not represent the event schemas of specific agent frameworks.
Future work
Potential extensions include:
- larger trace collections
- structured JSON trajectories
- LangGraph traces
- tool-call graphs
- multi-agent interactions
- human annotations
- learned failure-mode classifiers
- independently collected evaluation traces
License
Apache-2.0
- Downloads last month
- -