benchmark_slug stringclasses 6
values | bytes int64 1.25k 107k | description stringclasses 7
values | kind stringclasses 3
values | label stringclasses 7
values | path stringclasses 7
values | repositoryUrl stringlengths 88 104 | sha256 stringlengths 64 64 |
|---|---|---|---|---|---|---|---|
filingbench-50 | 32,768 | Pristine state copied for every episode. | SQLite | Seeded SQLite snapshot | environment.db | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/filingbench-50/environment.db | f70a5b84565d5aa98bbcb1aeff9c7a0742213f85c3ad876a46b0d59265875e51 |
filingbench-50 | 1,574 | Human-readable table and constraint definitions. | SQL | Relational schema | schema.sql | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/filingbench-50/schema.sql | 510bf7e2943495cb8a329c4ccce54ffc7376226dedd2b9c11288d8fa78dc40dd |
filingbench-50 | 14,116 | Human-readable source for the released world state. | SQL | Seed data | seed.sql | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/filingbench-50/seed.sql | c888acecc74c67054a759370315e8573f3315579c287ac540b21b57d37a3dc44 |
filingbench-50 | 104,507 | Public tool names, schemas, and implementations. | JSON | MCP tool catalog | tools.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/filingbench-50/tools.json | 6653f9d2a6e133047d7bb9f30b8ef81a74debb4efc94b94785664894036d4414 |
filingbench-50 | 106,953 | Instructions, splits, reference traces, and isolated verifier programs. | JSON | Task manifest | tasks.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/filingbench-50/tasks.json | 96388b0f39e51ed14f718001371783a506e5654586235e0ae1efd4decfbaf80b |
filingbench-50 | 2,039 | Stdio server declaration for the world. | JSON | MCP manifest | mcp/manifest.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/filingbench-50/mcp/manifest.json | e92914c7b33fd3e5a50cf7c908177962734d3e7e775da258ca4e62f1437fd301 |
filingbench-50 | 2,527 | Content identity, counts, split, and admission evidence. | JSON | World manifest | world.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/filingbench-50/world.json | 8475c86667f578b550dcc0bf5c449981036ea1a8c15dc9613906bb30165f972e |
servicebench-375 | 32,768 | Pristine state copied for every episode. | SQLite | Seeded SQLite snapshot | environment.db | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/servicebench-375/environment.db | f1c39d517436dd11eff05bf2da9440635d53b397ca3966e386e01f7f4a596537 |
servicebench-375 | 1,521 | Human-readable table and constraint definitions. | SQL | Relational schema | schema.sql | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/servicebench-375/schema.sql | aae113321a7e8003690d02fdf2e6540f74f9735211eaaefacbb8da160f481a2d |
servicebench-375 | 15,471 | Human-readable source for the released world state. | SQL | Seed data | seed.sql | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/servicebench-375/seed.sql | 9b5b097b96bcd595d8f3edc80d437e9881739f350848b4414840effa81dd7839 |
servicebench-375 | 101,198 | Public tool names, schemas, and implementations. | JSON | MCP tool catalog | tools.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/servicebench-375/tools.json | bbfffdac46d3bd61a8eea7e17c6c74effca6ceb2e2342daaf03cbb65f70da2a5 |
servicebench-375 | 105,899 | Instructions, splits, reference traces, and isolated verifier programs. | JSON | Task manifest | tasks.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/servicebench-375/tasks.json | 1a7d4d40cff99e0ae7e7fa21dc6fd5122f4f5a730fc939fec2c499e8046a613f |
servicebench-375 | 1,959 | Stdio server declaration for the world. | JSON | MCP manifest | mcp/manifest.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/servicebench-375/mcp/manifest.json | d7fc28450ed8676dec434717efb0d3e3b1b802ab58c4b904504d7b84192e83f5 |
servicebench-375 | 2,495 | Content identity, counts, split, and admission evidence. | JSON | World manifest | world.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/servicebench-375/world.json | 79dadcb06d047317b285f5f996c9e7568a386854d5e4f997b5667ce39bfda67c |
discoverybench-102 | 32,768 | Pristine state copied for every episode. | SQLite | Seeded SQLite snapshot | environment.db | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/discoverybench-102/environment.db | 44b907f65127934c49e2569a0226486c470231e635d324cec1aa09ebf4a6df05 |
discoverybench-102 | 1,497 | Human-readable table and constraint definitions. | SQL | Relational schema | schema.sql | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/discoverybench-102/schema.sql | 4dfc04a1e3d89b3839449644c07633a4bc2b86d65bc8d6d459ac0c3bb4989c83 |
discoverybench-102 | 11,964 | Human-readable source for the released world state. | SQL | Seed data | seed.sql | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/discoverybench-102/seed.sql | e3e3c20b66558ba03fff8da8292f92b28b031e1f2dc7a2310e5e702cff3398bc |
discoverybench-102 | 97,373 | Public tool names, schemas, and implementations. | JSON | MCP tool catalog | tools.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/discoverybench-102/tools.json | 511ab9dc9a947171a0a70b0d75d1201d40298c285f217b904ba69ef631ae5e5a |
discoverybench-102 | 103,570 | Instructions, splits, reference traces, and isolated verifier programs. | JSON | Task manifest | tasks.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/discoverybench-102/tasks.json | 73b32ea4a6834a886e55ae6cf26412bec6eadff9e06959912dbf905ab98881aa |
discoverybench-102 | 1,862 | Stdio server declaration for the world. | JSON | MCP manifest | mcp/manifest.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/discoverybench-102/mcp/manifest.json | 6dfe7eff3b010b684dadaac82edc606243c2582461bb691c512da2f3d7c074a4 |
discoverybench-102 | 2,492 | Content identity, counts, split, and admission evidence. | JSON | World manifest | world.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/discoverybench-102/world.json | 9788c5cbc28caa5c426e82ab59d220ad7d7aaccf913b4a77dc0e7a8e4f4c6f14 |
defensebench-100 | 32,768 | Pristine state copied for every episode. | SQLite | Seeded SQLite snapshot | environment.db | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/defensebench-100/environment.db | 4940300240cbe15d35d1fd45016398abe150ca2ee0361d04b2a17432f8a26233 |
defensebench-100 | 1,251 | Human-readable table and constraint definitions. | SQL | Relational schema | schema.sql | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/defensebench-100/schema.sql | 08c2e8abf194ee2764b390f01b9e8185c8653950c9582b3e2210e0f5f4a7b70b |
defensebench-100 | 14,152 | Human-readable source for the released world state. | SQL | Seed data | seed.sql | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/defensebench-100/seed.sql | 9c2b62927c80c430b0808605fbdefa36c200ef146cd539f6d02d6b1acb66573c |
defensebench-100 | 90,244 | Public tool names, schemas, and implementations. | JSON | MCP tool catalog | tools.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/defensebench-100/tools.json | 46021a577b0d1e7e3d8c5befed8ac5f6ea9d3d84ce61fc8e34a446fe59b35c43 |
defensebench-100 | 106,978 | Instructions, splits, reference traces, and isolated verifier programs. | JSON | Task manifest | tasks.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/defensebench-100/tasks.json | 44ae2da2ce9ef5faccb06335c0a242eabcdd8e41626925131bb01665d1165aa5 |
defensebench-100 | 1,969 | Stdio server declaration for the world. | JSON | MCP manifest | mcp/manifest.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/defensebench-100/mcp/manifest.json | 3c809b60c7ab96026e4ee0f4ac5ba0d04b011005465adff97185e0f98d1893ac |
defensebench-100 | 2,540 | Content identity, counts, split, and admission evidence. | JSON | World manifest | world.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/defensebench-100/world.json | c8a86f80f071188223d082706306fb84984c1fbb6cce9341f4ab0f8e4a48ef8a |
opsbench-14 | 32,768 | Pristine state copied for every episode. | SQLite | Seeded SQLite snapshot | environment.db | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/opsbench-14/environment.db | d6ca2df3416d0c441903d697c21418639839a44e6f43dd92cc0451baaead43fb |
opsbench-14 | 1,442 | Human-readable table and constraint definitions. | SQL | Relational schema | schema.sql | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/opsbench-14/schema.sql | 0e2ed2c463b00b1896d51e02a8abd3b9e743bfcdc73817ed8a09624f5fb8109d |
opsbench-14 | 13,353 | Human-readable source for the released world state. | SQL | Seed data | seed.sql | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/opsbench-14/seed.sql | 9457c0e86a30ba262563ebe9bbac22edeb38d07702018689154ec6e783373934 |
opsbench-14 | 99,191 | Public tool names, schemas, and implementations. | JSON | MCP tool catalog | tools.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/opsbench-14/tools.json | c8fbd5a4df7d08a3bdd0ad292f2cbc362f118e03f3c4b4c2da5f2ed7fbe0a315 |
opsbench-14 | 106,649 | Instructions, splits, reference traces, and isolated verifier programs. | JSON | Task manifest | tasks.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/opsbench-14/tasks.json | 0994054741ca6f7517a42584643e4570f434f7347dcf4fd59587f5525f97b1f8 |
opsbench-14 | 1,989 | Stdio server declaration for the world. | JSON | MCP manifest | mcp/manifest.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/opsbench-14/mcp/manifest.json | d431118d99a969b59a6f86bad4a178ddf673da1501091a11ba744a35df285dd5 |
opsbench-14 | 2,539 | Content identity, counts, split, and admission evidence. | JSON | World manifest | world.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/opsbench-14/world.json | 42447057cbe5a476888045d6601ad7c428d05a7aa3c2bc144cebb176bd85dd1b |
repobench-500 | 32,768 | Pristine state copied for every episode. | SQLite | Seeded SQLite snapshot | environment.db | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/repobench-500/environment.db | 12007f3262431078bc3f6476bec0de28d01e0b93acfcc2d86268d1ba36210303 |
repobench-500 | 1,690 | Human-readable table and constraint definitions. | SQL | Relational schema | schema.sql | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/repobench-500/schema.sql | 76b797e52cbe538d5907b14387a8da0e6fae3c883e546d5af1c1504a3973255f |
repobench-500 | 15,051 | Human-readable source for the released world state. | SQL | Seed data | seed.sql | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/repobench-500/seed.sql | d24ba1b70830c020f4abfa3a702a5e6a1bcb5378202a1a3c605af09fbcdecf47 |
repobench-500 | 103,816 | Public tool names, schemas, and implementations. | JSON | MCP tool catalog | tools.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/repobench-500/tools.json | 75077c064c2c026100e6e30008bac57e490f60918e72b48a0b8b18acfa1d61ea |
repobench-500 | 106,569 | Instructions, splits, reference traces, and isolated verifier programs. | JSON | Task manifest | tasks.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/repobench-500/tasks.json | c05032fc9010fed55dd7dd853f10622fa4657c7fe1844160270e8644bd26fda0 |
repobench-500 | 1,954 | Stdio server declaration for the world. | JSON | MCP manifest | mcp/manifest.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/repobench-500/mcp/manifest.json | 1b85dad89ae04a265b46ef7b53d52a8660c7973ca1158442637d84cf25de5faf |
repobench-500 | 2,517 | Content identity, counts, split, and admission evidence. | JSON | World manifest | world.json | https://github.com/blobfishai/blobfishai/tree/main/benchmark/worlds/repobench-500/world.json | c38c5e25cf4ba718e434d8da896a9c90ee02b5f8d25a6c9357c4ee10b2f3ff7d |
Blobfish DomainBench-24
DomainBench-24 is the public release record for 24 stateful agent tasks across six professional domains. Every task runs in a checked-in SQLite-backed MCP world with typed read/write tools and deterministic verification.
The release includes:
- 24 task records across six independent companion worlds;
- a 24/24 deterministic reference ceiling and a 0/24 seeded-random floor;
- 12 full sample trajectories: one successful reference replay and one failing floor episode per world;
- 42 hashed world assets, including the runnable database, schema, seed, tools, task manifest, MCP manifest, and world manifest.
- a digest-pinned Harbor release whose 24-task Oracle gate passed at reward 1.0 with zero exceptions.
No reference or random control is a model result. There are no frontier-model rankings in this release.
Domains and provenance
The Harbor datasets below are upstream professional-domain evaluation suites. The Blobfish worlds are independent diagnostic environments in the same domains; their 24 tasks are not converted or claimed as subsets of the upstream suites.
| Domain | Blobfish dataset page | Upstream Harbor source | Companion world |
|---|---|---|---|
| Financial research | FilingBench-50 | vals/financeagent | world |
| Customer operations | ServiceBench-375 | sierra-research/tau3-bench | world |
| Scientific discovery | DiscoveryBench-102 | scienceagentbench/scienceagentbench | world |
| Defensive security | DefenseBench-100 | polyvorlabs/cyberdefense-bench | world |
| Enterprise operations | OpsBench-14 | Enterprise-Bench/l1-l2-bench | world |
| Software engineering | RepoBench-500 | swe-bench/swe-bench-verified | world |
Files
data/tasks.jsonl— prompts, split labels, required tools, verifier basis, and control outcomes.data/trajectories.jsonl— stable trajectory metadata plus serialized tool steps and verifier results.data/assets.jsonl— byte size, SHA-256, and repository URL for each released asset.evidence/*.json— complete public evidence bundle used by blobfish.ai.worlds/<benchmark>/— the actual released world files.harbor-release.json— Harbor revision, task-package digests, runtime contract, and Oracle-gate receipt.MANIFEST.json— SHA-256 and byte length for every uploaded file.
Evaluation contract
Each episode starts from a fresh copy of the released SQLite state. The harness records tool calls, arguments, outputs, and errors, then grades the final state with deterministic VCode or an exact-answer check. Reference replay proves solvability; the seeded-random failure floor proves basic discrimination. Neither establishes empirical model difficulty.
License and upstream terms
Blobfish-authored release metadata and scripts follow the terms of the linked Blobfish repository. Upstream benchmark names, task sets, and source projects retain their own licenses. This release does not redistribute upstream Harbor tasks; consult each linked source page before using those separate datasets.
Reproduce
git clone https://github.com/blobfishai/blobfishai.git
cd blobfishai
node benchmark/validate.mjs
harbor run -d blobfishai/domainbench-24 -a <agent> -m <provider/model>
See the portfolio page for interactive task, asset, environment, trajectory, and methodology views.
- Downloads last month
- -