# Can Agent Benchmarks Support Their Scores? Evidence-Supported Bounds for Interactive-Agent Evaluation

arXiv: https://arxiv.org/abs/2605.10448 · HTML: https://mlsanity.com/paper/2605.10448/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Builds on 20 of its 39 references that are in the map.
- First posted on arXiv 11 May 2026.

## Builds on (most cited first, 20 of 20)

- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs](https://mlsanity.com/paper/2307.16789/) (2023)
- [τ-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains](https://mlsanity.com/paper/2406.12045/) (2024)
- [Datasheets for Datasets](https://mlsanity.com/paper/1803.09010/) (2018)
- [WebShop: Towards Scalable Real-World Web Interaction with Grounded Language Agents](https://mlsanity.com/paper/2207.01206/) (2022)
- [Mind2Web: Towards a Generalist Agent for the Web](https://mlsanity.com/paper/2306.06070/) (2023)
- [Model Cards for Model Reporting](https://mlsanity.com/paper/1810.03993/) (2018)
- [Dynabench: Rethinking Benchmarking in NLP](https://mlsanity.com/paper/2104.14337/) (2021)
- [AppWorld: A Controllable World of Apps and People for Benchmarking Interactive Coding Agents](https://mlsanity.com/paper/2407.18901/) (2024)
- [WorkArena: How Capable Are Web Agents at Solving Common Knowledge Work Tasks?](https://mlsanity.com/paper/2403.07718/) (2024)
- [TheAgentCompany: Benchmarking LLM Agents on Consequential Real World Tasks](https://mlsanity.com/paper/2412.14161/) (2024)
- [Improving Reproducibility in Machine Learning Research (A Report from the NeurIPS 2019 Reproducibility Program)](https://mlsanity.com/paper/2003.12206/) (2020)
- [ToolSandbox: A Stateful, Conversational, Interactive Evaluation Benchmark for LLM Tool Use Capabilities](https://mlsanity.com/paper/2408.04682/) (2024)
- [Reinforcement Learning on Web Interfaces Using Workflow-Guided Exploration](https://mlsanity.com/paper/1802.08802/) (2018)
- [Judging the Judges: A Systematic Study of Position Bias in LLM-as-a-Judge](https://mlsanity.com/paper/2406.07791/) (2024)
- [Rethinking Benchmark and Contamination for Language Models with Rephrased Samples](https://mlsanity.com/paper/2311.04850/) (2023)
- [AgentRewardBench: Evaluating Automatic Evaluations of Web Agent Trajectories](https://mlsanity.com/paper/2504.08942/) (2025)
- [Are "Solved Issues" in SWE-bench Really Solved Correctly? An Empirical Study](https://mlsanity.com/paper/2503.15223/) (2025)
- [SPHERE: An Evaluation Card for Human-AI Systems](https://mlsanity.com/paper/2504.07971/) (2025)
- [Is Your Benchmark Still Useful? Dynamic Benchmarking for Code Language Models](https://mlsanity.com/paper/2503.06643/) (2025)

## Built on by (most cited first, 5 of 5)

- [Do Agent Benchmarks Do What They Say? An Executable-Contract Audit of Tool-Using Agent Environments](https://mlsanity.com/paper/2609.37315/) (2026)
- [Success Is Not Self-Explanatory: Auditing Success Provenance in Agent Evaluation](https://mlsanity.com/paper/2607.24054/) (2026)
- [When Is an Agent Evaluation Over? Outcome Finality and Cross-Unit Separation](https://mlsanity.com/paper/2608.14940/) (2026)
- [Do Not Restart: Residual Completion for Stateful Agent Handoffs](https://mlsanity.com/paper/2609.13800/) (2026)
- [Get In, Decide, Prove: A Merchant-Side Readiness Benchmark for Agentic Commerce](https://mlsanity.com/paper/2610.17378/) (2026)

## Abstract (arXiv)

Interactive agent benchmarks map an agent run to a binary outcome through outcome checks. When these checks rely on surface level signals or fail to capture the agent's actual action path, they cannot reliably determine whether the run succeeded. For example, a benchmark task may ask whether Alice's shipping address was changed, while the outcome check only verifies that the agent clicked "Save." This does not guarantee that the intended state change occurred, since the agent may have modified the wrong record. Treating such a run as successful therefore makes the reported score misleading. Benchmark quality thus depends not only on task design, but also on the reliability of outcome detection. We address this problem by introducing an outcome evidence reporting layer for existing benchmarks, without modifying their tasks, agents, or evaluators. The layer performs three functions. First, before scoring, it specifies which stored artifacts are required to verify the claimed outcome for each case. Second, it applies a locked checklist to each completed run and assigns one of three evidence labels: Evidence Pass, Evidence Fail, or Unknown. Third, it reports evidence supported score bounds that quantify uncertainty arising from Unknown cases. Rather than silently counting, discarding, or hiding uncertain cases inside a single aggregate success rate, the framework keeps them explicitly visible. We evaluate the outcome evidence layer on five public benchmarks: ANDROIDWORLD, AGENTDOJO, APPWORLD, tau3 bench retail, and MINIWOB. The resulting reports separate several empirically distinct failure modes.