What it measures
Fact-seeking questions that remain difficult when search returns conflicting, noisy or unhelpful results. Seal-0 is the main challenge set; Seal-Hard and LongSeal probe additional reasoning and document-selection conditions.
Reading the measurement
The useful outcome is a verified change or deliverable. Inspect the tool interface, initial state, user simulator and policy constraints. An agent can claim completion without reaching the required state, or reach it through an action the protocol forbids.
Published evaluation
Seal-0 fact-seeking accuracy. Dagger-marked observations use built-in search; others use the paper FreshPrompt search setup. Search evidence can be noisy or contradictory.
Where this fits in an evaluation suite
A transcript that says “done” is weaker evidence than the verified final state. Track policy violations, recoveries, and invalid attempts alongside success.
Start with a capability hypothesis, then choose a second test with a different failure mode. Keep the checkpoint, reasoning effort, agent tools, sampling policy, and scoring procedure fixed when measuring a training intervention. Record failures and missing outputs in the denominator.
A reproducible comparison
Seal-0 fact-seeking accuracy. Dagger-marked observations use built-in search; others use the paper FreshPrompt search setup. Search evidence can be noisy or contradictory.
Publish the exact run configuration, item coverage, score units, and available uncertainty. If a source supplies only an aggregate or a single run, preserve that limit. A new model name or a rounded score alone does not establish a capability improvement.
PRIMARY EVIDENCE & FURTHER READING
- SealQA: Raising the Bar for Reasoning in Search-Augmented Language Models — published result table ↗
Reviewed 2026-10-26. Explanation and interpretation by RLTasks. Benchmark ownership remains with its original authors.