Prumo

A decision engine with zero learned parameters. It computes what can be computed and abstains, by construction, from everything else.

Model page: huggingface.co/tardellirs/prumo · Code: github.com/tardellirs/prumo

Prumo is Portuguese for plumb line: a tool that is exact without learning anything.

Prumo reads a state and a set of typed questions, the Decision Index / systemone wire format, and returns a probability for every option. It contains no model, no weights, no embeddings and no statistics fitted to a corpus. It is a fixed pipeline of generic capabilities, each of which recognises a structure in the request and solves it exactly:

• code to execute;
• a chess position;
• a catalogue of tools;
• candidates to rank;
• a logical expression;
• a date;
• an arithmetic problem.

When nothing applies, it says so with a uniform distribution.

Size is part of the result. Prumo sits at the zero-parameter end of the size/quality trade-off and puts a floor under it: this much of the public Decision Index is reachable without learning anything.

Highlights

• Full score 12.21 on the Decision Index 0.4 (maintainer-run private tests: same skills 15.78, new domains 5.60), above Crivo-17M (11.11), a learned 17M-parameter model.
• Public index 20.14 on the Decision Index 0.3 (shown on the board, not counted in 0.4), from a complete run of the official kit (140,178 requests). For reference, the other rule-based entrant scores 1.55, and learned models of 17M and 68M parameters score 7.44 and 13.01.
• Exact where it applies:
  • executes the code in CRUXEval (skill 0.997);
  • validates tool calls against their schemas (When2Call 0.70, BFCL 0.67);
  • searches chess positions;
  • solves algorithmic BIG-Bench Hard tasks.
• Honest elsewhere. Knowledge, commonsense and NLI benchmarks stay at chance by design. The engine never guesses beyond what it can compute.
• Fast and tiny:
  • median 0.2 ms per request on one CPU thread, mean 5.1 ms;
  • no GPU, no download beyond ~2.7k lines of Python and one pure-Python dependency.
• Deterministic:
  • every budget is counted in operations, not seconds;
  • the macOS and Linux runs give bit-identical answers.

What it is

Kind	rule-based, symbolic; 0 learned parameters
Size	527 hand-set numeric parameters, all in model.safetensors (the size the Hub reports); no weights, no embeddings
Input / output	Decision Index wire format: state + questions (choice, noul) → a probability for every option
Runtime	CPython ≥ 3.10, one CPU thread; dependency: chess==1.11.2 (move generation only)
Code	~2.7k lines (prumo_rules/), Apache-2.0
Language	English (lexical parts); the exact solvers are language-agnostic

How it decides

Every question goes through the same capabilities, in the same order, for every request:

code → chess → tools → retrieval → grounding → consensus → labels → music → logic → puzzles → dates → arith

Each capability either claims a question and returns probabilities, or passes it on. A question that no capability claims gets the uniform distribution, with a deterministic hash pick for the reported choice, so its expected score is exactly chance. The engine never looks at benchmark names, item ids or answer tables, and it uses one response format for every request.

Capability	Triggered by	What it computes
code	Python with a parseable def plus a call, or plus tests	runs it in a sandbox (below); picks the option equal to the value, or answers pass/fail
chess	a FEN string and legal-move options	alpha-beta with quiescence; material plus textbook piece-square tables; 7,000 nodes per question
tools	tool objects (name + description/schema)	BM25 relevance; for "which reply?" it validates the call against the schema (the tool exists, it is related to the request, every required value is user-stated) → call / ask / cannot
retrieval	a query plus candidates	BM25 with IDF over the request's own candidates; graded relevance from term coverage
grounding	a claim plus evidence, a response plus its source	lexical support: numbers and names must occur, and ≥ 70% of content words must be covered
consensus	participant opinions plus candidate statements	lexical centroid (mean normalised BM25 to every opinion)
labels	≥ 8 short label options	IDF-weighted overlap between the message and each label; out-of-scope wins when nothing overlaps
music	timed notes plus pitch-class options	chord-template matching (chord tones, non-chord tones, bass)
logic	Boolean expressions; premises plus a claim; liar/truth-teller puzzles	exact evaluation; truth tables (follows / contradicted / undetermined); SAT by enumeration
puzzles	ordering constraints, swaps, moves, schedules, SVG paths, brackets, word sorting	exhaustive search or simulation; claimed only when every sentence parses and the solution is unique
dates	an explicit anchor date plus date options	calendar arithmetic (days, weeks, months with end-of-month clamp, years, business days)
arith	an expression, or a word problem with numeric options	exact fractions; which option is reachable from the stated numbers (bounded dynamic programming)

Results

Decision Index 0.3 public suite, run once with this exact code through the official kit (pipeline --edition 0.3; complete: true, all 140,178 requests answered). In edition 0.4 the board shows this public index, but the Full score comes only from the maintainers' private tests (below).

Public index 20.14 (raw 38.96). Skill per area, chance-corrected:

Area	Skill
Tools & Automation	0.453
Retrieval & Classification	0.266
Knowledge & Reasoning	0.170
Arts & Human Taste	0.134
Language Understanding	0.031

The table below has every benchmark with skill above 0.03. Skill is chance-corrected: 0 is chance and 1 is perfect.

Benchmark	Skill	Capability
CRUXEval	0.997	code
When2Call	0.701	tools
BFCL	0.674	tools
POP909	0.520	music
BIG-Bench Hard	0.472	logic / puzzles / dates / arith
API-Bank	0.452	tools
CLINC150	0.447	labels
BANKING77	0.431	labels
ToolRet	0.394	retrieval
HoVer	0.393	grounding
ChessBench	0.333	chess
Habermas	0.246	consensus
RAGTruth	0.244	grounding
BRIGHT	0.204	retrieval
GSM8K	0.114	arith
Amazon ESCI	0.064	retrieval

Everything else (MMLU, MMLU-Pro, GPQA, ARC, HellaSwag, WinoGrande, ANLI, NLI4CT, VAST, CLadder, SATA-Bench, HLE, …) is at chance, as intended.

Public index of other submissions, from the board, for scale:

Submission	Learned parameters	Public index
Zev (rule-based)	0	1.55
Prumo	0	20.14
Bekko System One v0 17M	17M	7.44
Bekko System One v0 68M	68M	13.01
Bekko System One v0 400M	400M	22.45

Decision Index 0.4 — Full score (board)

Scored by the maintainers on the board in the Decision Index 0.4 batch of 2026-10-20. In edition 0.4 the Full score is 0.60 × private tests of the same skills + 0.40 × private tasks from new domains, both mapped onto the public scale using untuned stock models. The public index is still run and shown, but it no longer counts toward the Full score, and the new-domain tasks were revised. Edition 0.3.2 (0.20 public + 0.50 same skills + 0.30 new domains) stays available as an archive. Models within 0.9 points of the first model of a group share a rank group.

Full score 0.4	Private same skills	Private new domains	Public index (shown, not counted)	Board latency, median / p95 (1× RTX PRO 6000)	Rank group
12.21	15.78	5.60	20.14	0.3 / 11.5 ms	46 of 54

Rank group 46 is shared with Intern-Decision-0.8B (Full 12.67), Tev1-0.8B-experimental (Full 12.63) and Jev-Style 0.8B v3 (Full 12.07).

With zero learned parameters, Prumo scores above Crivo-17M (11.11, a 17M-parameter learned encoder). The board classes Prumo as an inference technique and, since it has no base model, lists no parameter count for it, so it does not appear on the board's size chart.

Compared with small and rule-based entries (board, Decision Index 0.4):

Model	Parameters	Full 0.4	Same skills	New domains
Prumo	not listed (0 learned)	12.21	15.78	5.60
Zev	not listed (0 learned)	1.91	1.79	1.89
Crivo-17M	16.9M	11.11	15.88	3.22
Bekko System One v0 17M	17.0M	5.72	5.89	4.61
Crivo-32M v2	32.0M	13.70	18.44	5.34
Bekko System One v0 68M	68.0M	9.99	12.62	4.98
Mola-68M	68.3M	14.96	21.99	3.52

On the private tests the exact capabilities transfer only in part: the same-skill score (15.78) is below the public index (20.14), and on the private tasks from new domains Prumo scores 5.60 (Zev, the other rule-based entry, 1.89).

Latency

One CPU thread, per request, over the whole public suite (Apple M2):

Median	Mean	p80	p99	Max
0.17 ms	5.1 ms	0.59 ms	132 ms	523 ms

The slowest paths are the chess search (median 118 ms, bounded by nodes) and the code sandbox (median 41 ms per start). The board's admission limit is 1,000 ms (median, mean or p80).

Determinism

Every budget is counted in operations, never in time:

• chess: 7,000 nodes;
• code: 10⁶ executed lines and calls;
• arithmetic: 60k operations.

The answers therefore do not depend on the machine. Two checks:

• Same machine, different processes: 1,100 requests re-run in a fresh process with a different PYTHONHASHSEED are bit-identical.
• Across operating systems: 3,370 requests (17,258 questions; 80 from each of the suite's 43 datasets) were run on Linux. That host was Ubuntu 24.04, x86-64, CPython 3.12.3, with no OS sandbox available. The answers are identical to the macOS run (Apple M2, CPython 3.12.13, sandbox-exec): agreement 1.0, maximum probability difference 0.0.

Usage

The Decision Index kit is not on PyPI; install it from its repository.

git clone https://github.com/apolinario/decision-index && pip install -e decision-index
git clone https://huggingface.co/tardellirs/prumo && cd prumo      # or github.com/tardellirs/prumo
pip install -r requirements.txt                                     # chess==1.11.2

# the public suite, exactly as submitted
python -m decision_index pipeline --engine decision_index_engine:DecisionEngine --edition 0.3 \
    --suite-dir <decision-index>/suite-0.3 --out runs/prumo

python -m unittest discover -s tests -v     # sandbox, determinism and latency checks (~5 s)

In Python:

from decision_index_engine import DecisionEngine

engine = DecisionEngine()

state = {"code": "def f(xs):\n    return [x * 2 for x in xs if x % 2]", "call": "f([1, 2, 3, 4, 5])"}
questions = {"output": {"type": "choice", "instructions": "What does the call return?",
                        "criteria": {"A": "[2, 6, 10]", "B": "[2, 4, 6, 8, 10]", "C": "[1, 3, 5]", "D": "[]"}}}
response, trace = engine(state, questions)
response["answers"]["output"]["choice"]   # "A"
trace                                     # {"solved_by": {"output": "code", "code_ms": 52.9}}

A question that no capability recognises comes back with uniform probabilities, and trace["solved_by"] stays empty for it. That is the signal to send the question to a learned model. Prumo works well as a zero-cost first stage in front of one.

To run without executing any code from the request, leave out the code capability:

--option solvers=chess,tools,retrieval,grounding,consensus,labels,music,logic,puzzles,dates,arith

Safety: executing code from requests

Code from a request never runs in the engine's own process. It runs in a child python -I -S -B, in a fresh empty temporary directory, with a minimal environment. The protections come in layers:

1. OS sandbox, when the host offers one:

   • macOS: sandbox-exec (no network, no writes outside the temp dir);
   • Linux: bwrap (all namespaces unshared, read-only root), or else unshare -rn.

   On hosts where unprivileged namespaces are blocked, the engine detects it and relies on layers 2–4. Ubuntu 24.04 with default AppArmor is one such host.

2. Resource limits: 1 s of CPU, 2 GiB of address space, 1 MiB per file, no new processes.

3. A CPython audit hook, installed before the code runs. It refuses:

   • sockets and network modules;
   • subprocess, fork, exec and spawn;
   • ctypes;
   • writes outside the temp dir;
   • reads outside the Python installation and the temp dir.

4. A deterministic budget of 10⁶ executed lines and calls, plus a 1 s wall-clock safety net.

A run that hits a limit counts as "cannot tell", and the question falls through. The tests in tests/test_engine.py check every layer on each host: no network, no writes, no reads of user files, no processes, budget and timeout. They pass on macOS and on Linux.

Transparency

• No learned values. Every number that shapes a decision is one of the 527 hand-set parameters in model.safetensors. The engine loads them at start-up (prumo_rules/params.py, standard library only, F64, so the values are bit-exact). tools/make_constants.py is their human-readable source, with a comment on each. The file holds:

  • 454 chess values: piece values plus seven textbook piece-square tables;
  • search constants;
  • the code, arithmetic and sandbox budgets;
  • BM25 k1 = 1.2, b = 0.75;
  • every threshold, weight, temperature and confidence mass.

  The file is named model.safetensors so that the Hub counts it, but it is not a trained model: it is the engine's configuration, kept where the size can be counted. Word lists (stop words, phrase lists, month and number words) and structural constants (board geometry, the calendar, the 12 pitch classes, text-length limits) stay in code. The thresholds that could not be fixed a priori were chosen on training/dev splits of the source datasets and on our own panels, never on the public test.

• Removed before release. An earlier prototype scored 24.81 on the public index, but it had parsers keyed to the fixed wording of particular benchmarks, and a few rules inspired by reading errors on the public test. All of them were removed:

  • the CLadder template parser;
  • the synthetic-phishing URL heuristics;
  • BIG-Bench Hard phrase anchors and word tables;
  • the dialogue boosts of API-Bank and When2Call.

  The tool rules were rebuilt from generic principles (schema validation, relatedness, recency) and validated on training data and on generated paraphrases. The public suite was then run once, for measurement only.

• Remaining caveat. A few constants of the prototype were first tried on public subsets before they were frozen: the chess evaluation, the chord templates, the arithmetic reachability budget and the consensus scoring. They are generic and kept, and they are declared here.

Limitations

• No world knowledge, commonsense or language understanding. MMLU-style, NLI, sentiment and stance tasks are at chance by design.
• Recognition is structural. A request whose structure the parsers do not recognise falls back to chance. Examples: code without a def, tools described only in prose, dates without an anchor. Prumo does not degrade gracefully toward partial understanding the way a learned model does.
• The lexical capabilities are heuristics (labels, grounding, retrieval, consensus). They help, but they are far from neural quality.
• Executing dataset code is a risk you should evaluate. Read the safety section, or disable code.

License

Apache-2.0 for this repository. The chess dependency (python-chess) is GPL-3.0 and is installed separately, not bundled.

Citation

@software{prumo2026,
  title  = {Prumo: a zero-parameter rule-based decision engine},
  author = {Tardelli Stekel},
  year   = {2026},
  url    = {https://huggingface.co/tardellirs/prumo}
}