whatmodel.sh / decision / Prumo (rule-based engine)

Prumo (rule-based engine)

+ contribute

Prumo (rule-based engine) is a source-available decision implementation from tardellirs. Licence: apache-2.0.

Catalogue identity notes

Decision Index entry prumo is a deterministic, symbolic decision engine with zero learned parameters. The pinned author README explicitly states that model.safetensors contains 527 hand-set numeric constants, not trained model weights. Retained as a rule-based comparison baseline; not a separately trained neural model. Historical benchmark scores and provenance are unchanged.

Model information sources

· Code licence ↗

22.7±1.1#335 of 355 decision modelsfrom 42 benchmark results across 1 source

$ how it compares

The decision models ranked either side of Prumo (rule-based engine), out of 357 with a score, then more from tardellirs.

#	model	score	where
335	MoganAI Lavoir ModernBERT-large Mogan AI	24±1	model files
336	GLiNER2.5-Decide Fastino	23±1	model files
337	Prumo (rule-based engine) tardellirs	23±1	source code
338	gliner2.5-small-v1 (DecisionBench 7e6f537f) Fastino	23±2	model files
339	Qwen-2.5-1B-RLCD (decision readout) harshatheg	21±1	source code
	more from tardellirs
253	Crivo-1B tardellirs	43±1	model files
279	Crivo-400M tardellirs	37±1	model files
301	Crivo-150M v2 tardellirs	32±1	model files

see all decision models · find the ones that fit your machine

$ what it's good at

tools & automation▁▃▁▁▁▃▃▄▆██▆ #212 of 250 · 10% of its score · 6 tests

Beats Cam1, InternLM Intern-Decision Qwen3.5-2B, Valen-0.8B here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: BFCL (case exact accuracy) submitted	75.9% ✎	▁▂▁▁▁▃▂▄▅▆█▁ beats 37/246	very easy · avg 92%	▇▇▇ 6.8%	sharp · at its level · noisy
Decision Index 0.4.1: API-Bank (accuracy) submitted	46.3% ✎	▁▁▁▁▁▁▁▃▃▂█▂ beats 42/236	easy · avg 70%	▇ 1.7%	sharp · noisy
Decision Index 0.4.1: When2Call (accuracy) submitted	77.5% ✎	▁▁▁▂▃▅▅▆█▇█▃ beats 129/246	easy · avg 72%	▇ 1.0%	at its level · noisy
Decision Index 0.4.1: ToolRet (nDCG@10) submitted	47.5% ✎	▁▁▁▁▁·▁▃▁▄█▁ beats 26/246	medium · avg 60%	▇ 0.7%	blunt · at its level
Decision Index 0.4.1: Home appliance simulator (case exact accuracy) submitted	0.0% ✎	█▁▁▃▄▄▆▄▁▂▁▃ beats 9/243	medium · avg 37%	▇ 0.1%	sharp · too hard for it · noisy
Decision Index 0.4.1: RouterBench (selected quality (quality objective)) submitted	57.9% ✎	no other model yet		—	failed run: Publisher excludes this protocol because its prompt reveals GPT-4 as the best route. Evidence: https://huggingface.co/spaces/multimodalart/jev-decision-index/blob/c130faf65f82e81d3bdefb01e758c712dc5c5199/data/index.json

retrieval & classification▁▁▁·▁▃▁▁▃▃█▅ #239 of 258 · 4.5% of its score · 7 tests

Beats Sieve 2B, DM-JEPA 1.1, Together Tev1 Qwen3.5-0.8B here, though they rank higher overall.

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: Amazon ESCI (macro-F1) submitted	25.4% ✎	▁▁▁▁▁▁▃▂▃▃█▅ beats 19/246	medium · avg 47%	▇ 1.2%	at its level
Decision Index 0.4.1: BRIGHT (nDCG@10) submitted	29.6% ✎	▁▁▁▁▁▁▁▃▂▄▇█ beats 48/246	medium · avg 38%	▇ 1.2%	blunt
Decision Index 0.4.1: CLINC150+OOS (macro-F1) submitted	45.0% ✎	▁▁▁▁▁▁▃▂▃▆█▄ beats 24/238	easy · avg 83%	▇ 0.6%	sharp · at its level · noisy
Decision Index 0.4.1: HoVer (accuracy) submitted	69.7% ✎	▆▆▇▅▄▄▃▇▄█▆▄ beats 97/246	easy · avg 74%	▇ 0.6%	blunt · at its level
Decision Index 0.4.1: PhishNChips phishing decisions (accuracy) submitted	50.0% ✎	█▅▅▄▃▂▁▁▁▁·▁ beats 11/245	easy · avg 70%	▇ 0.5%	at its level · noisy
Decision Index 0.4.1: BANKING77 (macro-F1) submitted	43.8% ✎	▁·▁▁▁▁▁▃▂█▇█ beats 14/238	easy · avg 79%	▇ 0.4%	at its level · noisy
Decision Index 0.4.1: SGD/SGD-X (macro-F1) submitted	23.2% ✎	no other model yet		—	failed run: Publisher reports a builder bug exposing dialogue turns after the classified turn; awaiting corrected rerun. Evidence: https://huggingface.co/spaces/multimodalart/jev-decision-index/blob/c130faf65f82e81d3bdefb01e758c712dc5c5199/data/index.json

language understanding▃▂▂▃▂▅▆▃█▆▄▃ #256 of 256 · 23% of its score · 9 tests

Trails Zev (decision readout), LFM2.5-350M-RLCD, Nirnaya 0.4B here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: HellaSwag (accuracy) submitted	24.8% ✎	▃▄▂▂▄▇▃▇█▅▃▁ beats 0/246	very easy · avg 85%	▇▇▇▇ 9.3%	sharp · at its level
Decision Index 0.4.1: NLI4CT (macro-F1) submitted	40.0% ✎	▁▁▃▂▂▃▃▄▇▇█▄ beats 5/246	easy · avg 74%	▇▇ 5.5%	at its level
Decision Index 0.4.1: ANLI (macro-F1) submitted	16.7% ✎	▃▂▃▅▅█▂····▁ beats 2/246	medium · avg 60%	▇ 3.5%	at its level
Decision Index 0.4.1: FinEntity (macro-F1) submitted	32.4% ✎	▁···▁▁▃▁▃▆█▃ beats 6/246	easy · avg 84%	▇ 1.0%	at its level · noisy
Decision Index 0.4.1: ContractNLI (macro-F1) submitted	17.5% ✎	▁▁▃▁▁▁▁▂▂▇▇█ beats 3/246	easy · avg 71%	▇ 1.0%	at its level · noisy
Decision Index 0.4.1: iSarcasmEval (Sarcasm F1 · track A, English) submitted	25.0% ✎	▁·▁·▁▁▂▄▅█▇▄ beats 35/246	medium · avg 44%	▇ 1.0%	typical
Decision Index 0.4.1: VAST (macro-F1) submitted	16.9% ✎	▃▁▁▂▅▄▄▂▄▆█▄ beats 1/246	medium · avg 62%	▇ 0.7%	at its level
Decision Index 0.4.1: RAGTruth (F1 on hallucinated class) submitted	63.5% ✎	▁▁▁▁▁▁▁▁▁▃▆█ beats 74/246	easy · avg 66%	▇ 0.6%	at its level · noisy

1 more test

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: ACOS (per-review F1) submitted	3.9% ✎	▁··▂▃▃▅▇█▇▃▃ beats 17/246	hard · avg 22%	▇ 0.2%	too hard for it · noisy

arts & human taste▁▇▄▃▅▅█▅▄▂▁▁ #231 of 247 · 7.0% of its score · 6 tests

Beats JEV Ettin 150M, JEV FLAN-T5 Large public v2 (342M), Kev 0.8B r15 here, though they rank higher overall. Trails LFM2.5-350M-RLCD, system-one-gemma, Bekko System One v0 17M here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: New Yorker (accuracy) submitted	18.4% ✎	▃▁▂▃▃▃▇▅█▅▃▁ beats 0/246	medium · avg 61%	▇ 2.7%	at its level
Decision Index 0.4.1: BPoMP (accuracy) submitted	50.3% ✎	▁█▄▄▅█▆▅▄▁▁▁ beats 16/246	easy · avg 83%	▇ 1.9%	sharp · at its level · noisy
Decision Index 0.4.1: Humicroedit (accuracy) submitted	50.9% ✎	▁·▁▃▂▂▅█▄▂▁▁ beats 28/246	medium · avg 60%	▇ 1.4%	blunt · at its level
Decision Index 0.4.1: cfcolor (accuracy) submitted	50.2% ✎	▁········▁▇█ beats 19/246	medium · avg 60%	▇ 0.8%	blunt · at its level
Decision Index 0.4.1: Habermas Machine (accuracy) submitted	48.0% ✎	▁▂▄█▇▅▁▁▄▄▁▁ beats 168/246	medium · avg 47%	▇ 0.2%	blunt · at its level
Decision Index 0.4.1: POP909-CL (accuracy) submitted	53.0% ✎	▄▅▅▃▆▅▃▅▂▅▃█ beats 168/237	hard · avg 21%	▇ 0.1%	sharp · too hard for it · noisy

knowledge & reasoning▃▂▂▂▂▃▃▅▅▆█▂ #260 of 263 · 49% of its score · 14 tests

Trails Supersonic Labs Julia 1, system-one-gemma, Bekko System One v0 17M here, though they rank lower overall.

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: ARC-Easy (accuracy) submitted	24.7% ✎	▃▁▂▁▂▂▂▂▃▆█▅ beats 0/246	very easy · avg 96%	▇▇▇▇▇▇ 15%	sharp · at its level · noisy
Decision Index 0.4.1: ARC-Challenge (accuracy) submitted	26.5% ✎	▃▂▂▂▂▂▁▃▄▆█▆ beats 4/246	very easy · avg 90%	▇▇▇▇ 10%	sharp · at its level
Decision Index 0.4.1: MMLU (accuracy) submitted	24.7% ✎	▃▄▄▃▃▂▅▇▅█▇▁ beats 3/246	easy · avg 70%	▇▇ 5.9%	sharp · at its level
Decision Index 0.4.1: BBH (accuracy) submitted	63.6% ✎	▄▆▄▄█▇▇▄▃▁▁▁ beats 99/246	easy · avg 66%	▇▇ 4.7%	at its level
Decision Index 0.4.1: WinoGrande (accuracy) submitted	50.4% ✎	█▅█▅▄█▆▄▃▃▁▁ beats 7/246	easy · avg 78%	▇ 3.7%	at its level
Decision Index 0.4.1: MMLU-Pro (accuracy) submitted	11.1% ✎	▄▄▃▃▂▄▅▇██▅▁ beats 1/246	medium · avg 46%	▇ 3.6%	sharp
Decision Index 0.4.1: GPQA Diamond (accuracy) submitted	24.0% ✎	▂▆▅▅▆█▆▃▁··▁ beats 11/246	medium · avg 39%	▇ 1.6%	blunt
Decision Index 0.4.1: GSM8K (accuracy (4- and 10-choice)) submitted	26.7% ✎	▁··▄█▆▆▆▅▅▃▁ beats 64/246	medium · avg 45%	▇ 1.1%	sharp

6 more tests

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: MuSR (accuracy) submitted	36.0% ✎	▃▁▂█▅▄▁▁▁▁▁▁ beats 3/246	medium · avg 60%	▇ 0.9%	blunt · at its level
Decision Index 0.4.1: CLadder (accuracy) submitted	50.1% ✎	▁███▆▅▃▃▁▁▁▁ beats 10/246	easy · avg 71%	▇ 0.9%	at its level
Decision Index 0.4.1: CRUXEval (accuracy) submitted	99.8% ✎	██▄▃▃▄▁▁···▃ beats 246/246	medium · avg 58%	▇ 0.7%	at its level · noisy
Decision Index 0.4.1: SATA-Bench (case exact accuracy) submitted	2.9% ✎	▁▁▁▃▁▁▁▂▅█▂▁ beats 24/246	hard · avg 22%	▇ 0.4%	too hard for it
Decision Index 0.4.1: ChessBench (accuracy) submitted	38.7% ✎	▄▇▄▄▄▆▄█▄▁▁▃ beats 245/246	hard · avg 16%	▇ 0.2%	blunt
Decision Index 0.4.1: HLE (accuracy) submitted	14.2% ✎	▁▁▂▃▆█▇▅▅▃▁▁ beats 178/246	very hard · avg 13%	▇ 0.0%	blunt

other tests2 not in a task · 5.8% of its score

test	result	vs the others in its run	difficulty	counts	why
Decision Index 0.4.1: private same skills (equated points)	14.9100 points ✎	▂▂▃▃▃▃▄▅▅▇█▃ beats 25/247	avg 41.206 points	▇ 3.7%	blunt
Decision Index 0.4.1: private new domains (equated points)	12.6200 points ✎	▃█▃▃▂▆▃▃▄▄▄▄ beats 39/247	avg 38.861 points	▇ 2.1%	blunt

$ where to run it

Implementation and setup instructions ↗
Published Python rule engine and setup documentation verified. The 10,320-byte safetensors file stores hand-set constants; this is source-code access, not learned weights. No local execution performed during this review. Source · checked 2026-11-01.

Model repository and documentation: huggingface.co/tardellirs ↗

$ how the score was worked out · Decision Index 0.4.1 100%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

source	results	share of the score	licence
Decision Index 0.4.1 · submitted; independent execution unverified	42	▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 100%	CC BY 4.0
prior · pull toward the middle		▇ 0%
weighted average		= 22.7

not counted (2)

Decision Index 0.4.1	Decision Index 0.4.1: RouterBench (selected quality (quality objective)) [correct]	57.9%	failed run: Publisher excludes this protocol because its prompt reveals GPT-4 as the best route. Evidence: https://huggingface.co/spaces/multimodalart/jev-decision-index/blob/c130faf65f82e81d3bdefb01e758c712dc5c5199/data/index.json
Decision Index 0.4.1	Decision Index 0.4.1: SGD/SGD-X (macro-F1) [correct]	23.2%	failed run: Publisher reports a builder bug exposing dialogue turns after the classified turn; awaiting corrected rerun. Evidence: https://huggingface.co/spaces/multimodalart/jev-decision-index/blob/c130faf65f82e81d3bdefb01e758c712dc5c5199/data/index.json

$ improve this page

Update model information Add a place to run it Add a test result Propose a benchmark Correct a benchmark score Edit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated 2026-11-03 02:37 UTC · sources and licences · json