context_files listlengths 28 28 | gold_output dict | metadata dict | prompt stringlengths 369 491 | rubric dict | task_id stringlengths 26 37 | task_name stringlengths 25 45 | world_id stringclasses 1
value |
|---|---|---|---|---|---|---|---|
[
"task_files/sb100-001-northwind-q3-commit/01_salesforce_pipeline/01_01_salesforce_pipeline_register.md",
"task_files/sb100-001-northwind-q3-commit/02_hubspot_deals/02_02_hubspot_deals_register.csv",
"task_files/sb100-001-northwind-q3-commit/03_gong_briefs/03_03_gong_briefs_register.json",
"task_files/sb100-00... | {
"brief_markdown": "# Northwind Q3 commit inspection\n\n## Executive assessment\n\nThe board forecast includes channel deals whose Salesforce stages moved forward while HubSpot still shows unsigned evaluations; Gong records reveal which buyers actually confirmed procurement dates. The evidence supports 7 bounded cha... | {
"benchmark": "SalesBench-100",
"code_license": "Apache-2.0",
"company": "Northwind Grid Systems",
"contract_pins": {
"filesystem": {
"commit": "9a96ea6e5913736f92b88345bf51caeaaa8e719f",
"package": "@modelcontextprotocol/server-filesystem",
"repository": "https://github.com/modelcontextp... | Can you get northwind q3 commit inspection settled for VP of Revenue Operations before the 2026-Q3 review? The board forecast includes channel deals whose Salesforce stages moved forward while HubSpot still shows unsigned evaluations; Gong records reveal which buyers actually confirmed procurement dates. Make the chang... | {
"authorized_mutations": 7,
"call_order_policy": "The reference trajectory is illustrative, not graded. Material identity, authority, provider, calendar, and exception evidence must precede the dependent CRM mutation; valid query shapes and the order of independent investigations are open.",
"criteria": [
{
... | sb100-001-northwind-q3-commit | Northwind Q3 commit inspection | salesbench-multi-crm-mcp-v3 |
["task_files/sb100-002-velora-emea-rollup/01_salesforce_pipeline/01_01_salesforce_pipeline_register.(...TRUNCATED) | {"brief_markdown":"# Velora EMEA rollup repair\n\n## Executive assessment\n\nA regional rollup doubl(...TRUNCATED) | {"benchmark":"SalesBench-100","code_license":"Apache-2.0","company":"Velora Mobility","contract_pins(...TRUNCATED) | "EMEA Revenue Operations Lead has asked for a clean answer on velora emea rollup repair. A regional (...TRUNCATED) | {"authorized_mutations":11,"call_order_policy":"The reference trajectory is illustrative, not graded(...TRUNCATED) | sb100-002-velora-emea-rollup | Velora EMEA rollup repair | salesbench-multi-crm-mcp-v3 |
["task_files/sb100-003-harborstone-public-sector/01_salesforce_pipeline/01_01_salesforce_pipeline_re(...TRUNCATED) | {"brief_markdown":"# Harborstone public-sector forecast audit\n\n## Executive assessment\n\nFederal (...TRUNCATED) | {"benchmark":"SalesBench-100","code_license":"Apache-2.0","company":"Harborstone Cloud","contract_pi(...TRUNCATED) | "We need to finish harborstone public-sector forecast audit for Harborstone Cloud this cycle. Federa(...TRUNCATED) | {"authorized_mutations":10,"call_order_policy":"The reference trajectory is illustrative, not graded(...TRUNCATED) | sb100-003-harborstone-public-sector | Harborstone public-sector forecast audit | salesbench-multi-crm-mcp-v3 |
["task_files/sb100-004-cedarline-midmarket/01_salesforce_pipeline/01_01_salesforce_pipeline_register(...TRUNCATED) | {"brief_markdown":"# Cedarline mid-market forecast scrub\n\n## Executive assessment\n\nA manager bul(...TRUNCATED) | {"benchmark":"SalesBench-100","code_license":"Apache-2.0","company":"Cedarline Analytics","contract_(...TRUNCATED) | "Could you take ownership of cedarline mid-market forecast scrub for Cedarline Analytics? A manager (...TRUNCATED) | {"authorized_mutations":6,"call_order_policy":"The reference trajectory is illustrative, not graded.(...TRUNCATED) | sb100-004-cedarline-midmarket | Cedarline mid-market forecast scrub | salesbench-multi-crm-mcp-v3 |
["task_files/sb100-005-atlas-apac-currency/01_salesforce_pipeline/01_01_salesforce_pipeline_register(...TRUNCATED) | {"brief_markdown":"# Atlas APAC currency-normalized forecast\n\n## Executive assessment\n\nHubSpot s(...TRUNCATED) | {"benchmark":"SalesBench-100","code_license":"Apache-2.0","company":"Atlas Robotics","contract_pins"(...TRUNCATED) | "Before Atlas Robotics's 2026-Q3 review, APAC Finance Business Partner needs atlas apac currency-nor(...TRUNCATED) | {"authorized_mutations":9,"call_order_policy":"The reference trajectory is illustrative, not graded.(...TRUNCATED) | sb100-005-atlas-apac-currency | Atlas APAC currency-normalized forecast | salesbench-multi-crm-mcp-v3 |
["task_files/sb100-006-lattice-partner-overlay/01_salesforce_pipeline/01_01_salesforce_pipeline_regi(...TRUNCATED) | {"brief_markdown":"# Lattice partner-overlay de-duplication\n\n## Executive assessment\n\nPartner-so(...TRUNCATED) | {"benchmark":"SalesBench-100","code_license":"Apache-2.0","company":"Lattice Security","contract_pin(...TRUNCATED) | "The next operating meeting is blocked on lattice partner-overlay de-duplication. Partner-sourced op(...TRUNCATED) | {"authorized_mutations":7,"call_order_policy":"The reference trajectory is illustrative, not graded.(...TRUNCATED) | sb100-006-lattice-partner-overlay | Lattice partner-overlay de-duplication | salesbench-multi-crm-mcp-v3 |
["task_files/sb100-007-meridian-usage-expansion/01_salesforce_pipeline/01_01_salesforce_pipeline_reg(...TRUNCATED) | {"brief_markdown":"# Meridian usage-expansion outlook\n\n## Executive assessment\n\nExpansion opport(...TRUNCATED) | {"benchmark":"SalesBench-100","code_license":"Apache-2.0","company":"Meridian Data Fabric","contract(...TRUNCATED) | "Meridian Data Fabric needs a reliable result for meridian usage-expansion outlook, not another dash(...TRUNCATED) | {"authorized_mutations":5,"call_order_policy":"The reference trajectory is illustrative, not graded.(...TRUNCATED) | sb100-007-meridian-usage-expansion | Meridian usage-expansion outlook | salesbench-multi-crm-mcp-v3 |
["task_files/sb100-008-opal-healthcare-commit/01_salesforce_pipeline/01_01_salesforce_pipeline_regis(...TRUNCATED) | {"brief_markdown":"# Opal healthcare commit certification\n\n## Executive assessment\n\nHospital dea(...TRUNCATED) | {"benchmark":"SalesBench-100","code_license":"Apache-2.0","company":"Opal Clinical Network","contrac(...TRUNCATED) | "Please sort out opal healthcare commit certification for Healthcare Sales VP. Hospital deals cross (...TRUNCATED) | {"authorized_mutations":5,"call_order_policy":"The reference trajectory is illustrative, not graded.(...TRUNCATED) | sb100-008-opal-healthcare-commit | Opal healthcare commit certification | salesbench-multi-crm-mcp-v3 |
["task_files/sb100-009-quarry-enterprise-split/01_salesforce_pipeline/01_01_salesforce_pipeline_regi(...TRUNCATED) | {"brief_markdown":"# Quarry enterprise split-credit reconciliation\n\n## Executive assessment\n\nPar(...TRUNCATED) | {"benchmark":"SalesBench-100","code_license":"Apache-2.0","company":"Quarry Compute","contract_pins"(...TRUNCATED) | "There is a live operating decision behind quarry enterprise split-credit reconciliation at Quarry C(...TRUNCATED) | {"authorized_mutations":11,"call_order_policy":"The reference trajectory is illustrative, not graded(...TRUNCATED) | sb100-009-quarry-enterprise-split | Quarry enterprise split-credit reconciliation | salesbench-multi-crm-mcp-v3 |
["task_files/sb100-010-solstice-fiscal-boundary/01_salesforce_pipeline/01_01_salesforce_pipeline_reg(...TRUNCATED) | {"brief_markdown":"# Solstice fiscal-boundary forecast reset\n\n## Executive assessment\n\nThe compa(...TRUNCATED) | {"benchmark":"SalesBench-100","code_license":"Apache-2.0","company":"Solstice Learning","contract_pi(...TRUNCATED) | "Can you close the loop on solstice fiscal-boundary forecast reset? The company changed fiscal calen(...TRUNCATED) | {"authorized_mutations":9,"call_order_policy":"The reference trajectory is illustrative, not graded.(...TRUNCATED) | sb100-010-solstice-fiscal-boundary | Solstice fiscal-boundary forecast reset | salesbench-multi-crm-mcp-v3 |
End of preview. Expand in Data Studio
SalesBench-100
SalesBench-100 is a synthetic long-horizon sales-agent benchmark with 100 original workflows across Salesforce, HubSpot, Gong, and a seeded evidence room. Each task begins with a high-level employee request and has its own authored causal rule and provider transition. Identity, operating facts, authority, governed policy, live-system indexes, and exceptions are separated so no mounted business asset publishes a selected option or precomputed change. Every task has 28 independently inspectable assets across 11 native formats; 16–18 task-specific records are causally material and the remainder provide realistic corroboration, decoys, lineage, and surrounding context. The evidence—not a fixed mutation quota—determines 5–12 authorized CRM mutations and 68–114 reference calls, including a same-record post-write readback for every mutation. Every task also grades a decision model: three alternatives (standard queue, expedited exception queue, full hold) each carrying an outcome date derived across the business-day calendar, a whole-USD incremental cost, and an authority status, compared with a review meeting date documented in two independent sources into a signed variance and an ON_TIME/LATE status.
Public release
- Runnable Harbor world: https://hub.harborframework.com/datasets/blobfishai/salesbench-100
- Dataset and test assets: https://huggingface.co/datasets/SamuelChien821/salesbench-100
- Benchmark page: https://blobfish.ai/benchmarks/salesbench-100
- Builder and verifier source: https://github.com/blobfishai/sales-agent-simulation
Contents
data/tasks.jsonl: portable task records with prompts, context paths, deterministic rubric metadata, and public gold output.tasks/: one readable JSON record per workflow.task_files/: 2,800 unique task-scoped assets across PDF, Excel, email, Slack/Drive-style JSON, structured exports, controls, and audit records.world/: the four-surface offline MCP world and verifier.contracts/: pinned contract metadata and exact published tool schemas.trajectories/: accepted tool traces produced by the qualification run.reports/: build, qualification, MCP-conformance, and container-probe evidence.
Objective release gates
| Gate | Required |
|---|---|
| Tasks | 100 |
| Workflow families | 10 |
| Agent-visible assets per task | 28 across 11 native formats |
| Authorized mutations per task | 5–12; at least six distinct workload sizes |
| Reference MCP calls per task | 75–114 |
| Unique tool-name sequences | 100/100 |
| Unique semantic action graphs | 100/100 |
| Authored causal decision rules | 100/100 unique |
| Task-specific semantic milestones | 15 totaling 100 points; every atomic criterion is published with its milestone |
| Costed, dated, authority-tagged alternatives | 3 per task; one APPROVED, one ADDITIONAL_APPROVAL_REQUIRED, one AVAILABLE_NOT_RECOMMENDED |
| Business-need date documented outside the request | 2 independent sources per task; absent from every prompt |
| Approval-pending and superseded-period holds | at least one of each per task |
| Precomputed answer objects or option outcomes in business evidence | 0 |
| Material evidence and provider reads before mutation | required; independent order and valid query shapes are open |
| Same-record post-write readback per mutation | required |
| Verifier network/model/clock/random calls | 0 |
| Oracle and replay passes | 100/100 each |
| False accepts across thirteen negative controls | 0 |
| Reward for the pristine no-op control | exactly 0.0 |
Measured results live in reports/qualification.json. Reference traces are implementation proofs, not model scores.
Contract fidelity
The world keeps Salesforce, HubSpot, Gong, and filesystem as separate MCP endpoints. Names and input JSON Schemas are pinned to official hosted documentation or real open-source implementations at immutable commits recorded in contracts/upstream-pins.json. The Gong surface is read-only. Output envelopes mirror the corresponding MCP/API JSON shapes, with documented offline simplifications.
Data and licensing
All people, entities, records, messages, amounts, and commercial facts are synthetic. CRMArena, SCUBA, AutomationBench, Harvey Labs, and Apex Accounting informed quality and packaging analysis only; their task text and data are not included. Synthetic task data is CC-BY-4.0; generator/runtime code is Apache-2.0. Third-party contract facts retain their upstream terms.
- Downloads last month
- 1,848