Provider and attribution
LMArena Leaderboard Dataset
- Registry state
- registered
- Rights state
- approved
- Provider group
- lmarena
Contains data from the Arena Leaderboard Dataset by Arena, licensed under CC BY 4.0
Release planning · no model scores
Agent outcome and steerability evidence mapping
Exact V3 collection, facet, metric, lineage, identity and rights mapping for Agent outcome and steerability from LMArena Leaderboard Dataset.
V3 evidence records
1 collection mapping
A repeated upstream benchmark can map to several collections. Each record below preserves its own task role, facets, capability atoms, subject identity, metrics, lineages and exclusions.
supporting evidence · business-automation/lmarena/agent-outcome-and-steerability
Business Workflow Automation
Which models are best at business workflow automation?
- Upstream access
- machine-ready
- Eligibility
- runnable-now
- Registry
- registered
- Max task directness
- 45%
Collection facets
Workflow completion · objective · 25%
Reach the correct end state across multiple dependent steps.
Exception recovery · hybrid · 25%
Recover from missing data, failed tools, and conflicting state.
Approval and action safety · hybrid · 25%
Pause for authorization before consequential or irreversible actions.
Operational communication · hybrid · 25%
Explain progress, blockers, and handoffs without excess narration.
Capability atoms
task_outcomesteerabilitytool_grounding
Subject identity
- Declared:
- harnessed-model
- Metric partitions:
- harnessed-model
- Separate partition required:
- no
- Identity projection:
- forbidden
Usable evidence lineages
lmarena-agent-causal-tracing/harnessed-model
Exact metric contract
| Metric | Lineage | Subject | Protocol | Use |
|---|---|---|---|---|
agent.task_outcome_explicitLMArena Agent — Confirmed success net improvement | lmarena-agent-causal-tracing | harnessed-model | Agent Arena causal treatment effect for explicit task outcome | usable |
agent.steerabilityLMArena Agent — Steerability net improvement | lmarena-agent-causal-tracing | harnessed-model | Agent Arena causal treatment effect for accepted user corrections | usable |
agent.tool_hallucinationLMArena Agent — Tool-hallucination outcome net improvement | lmarena-agent-causal-tracing | harnessed-model | Agent Arena causal treatment effect for avoiding nonexistent tools | usable |
Exclusions and blockers
No mapping-level exclusion is recorded; the page-level release gate still applies.
Mapping note
Agent arena results retain their harness identity.