SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2604.22893 · 2026

Utility-Aware Data Pricing: Token-Level Quality and Empirical Training Gain for LLMs

Kun Li, Qi Luo, Minghui Xu

arXiv · PDF · Open in the Atlas

Code that ran

6 of 8 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (1 of them constructed an object rather than computing a result; 2 honoured the contract we drafted, 1 violated it and 1 has no contract verdict; 2 failures of our instrument; 2 no recorded run), measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2604.22893.json · badge: /api/ran/2604.22893.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
BDS-SDU/utility-aware-data-pricing community
Papers with Code archive
6 ran · 2 no recorded run
FunctionStatusWhere it lives
DocumentRecord Ran
ran on a synthesized input · constructed an object; no result was computed
BDS-SDU/utility-aware-data-pricing/data_pricing/valuation.py
code served (permissive licence) · get_code("cc201e40557348b1")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned DocumentRecord: the class was constructed and called, and the call returned an instance of its own class. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

from typing import Any, Dict
from dataclasses import dataclass, field

@dataclass
class DocumentRecord:
    """Canonical training or validation sample."""
    doc_id: str
    source_id: str
    text: str
    label: int = 0
    metadata: Dict[str, Any] = field(default_factory=dict)

    @classmethod
    def from_dict(cls, payload: Dict[str, Any]) -> 'DocumentRecord':
        return cls(doc_id=str(payload['doc_id']), source_id=str(payload['source_id']), text=str(payload['text']), label=int(payload.get('label', 0)), metadata=dict(payload.get('metadata', {})))

    def to_dict(self) -> Dict[str, Any]:
        return {'doc_id': self.doc_id, 'source_id': self.source_id, 'text': self.text, 'label': self.label, 'metadata': self.metadata}


INIT_ARGS = {
    'doc_id': 'doc_001',
    'source_id': 'src_001',
    'text': 'This is a sample document text.',
    'label': 1,
    'metadata': {'author': 'John Doe', 'date': '2024-01-01'}
}

ARGS = {
    'payload': {
        'doc_id': 'doc_002',
        'source_id': 'src_002',
        'text': 'Another sample document.',
        'label': 0,
        'metadata': {'category': 'news'}
    }
}

METHOD = 'from_dict'

# Construction and execution
record = DocumentRecord(**INIT_ARGS)
print("Constructed record:", record)
print("Record as dict:", record.to_dict())

# Call the computation method
new_record = DocumentRecord.from_dict(**ARGS)
print("Record from dict:", new_record)
print("New record as dict:", new_record.to_dict())
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DocumentRecord — the call returned an instance of its own class, which has no array shape; none is stated — the instrument coerces any output to an array; for a DocumentRecord output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
conjugate_gradient Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
BDS-SDU/utility-aware-data-pricing/data_pricing/valuation.py
code served (permissive licence) · get_code("b53e5c166ea9d86c")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import math
from typing import Callable, Sequence, Tuple, List

# Define a simple matvec function that takes (w, b) and returns (result_w, result_b)
def simple_matvec(w: Sequence[float], b: float) -> Tuple[List[float], float]:
    """
    A simple matrix-vector product stand-in.
    Multiplies the input by a small matrix and bias.
    """
    # Simple operation: scale by 2 and add a small constant
    result_w = [2.0 * val for val in w]
    result_b = 2.0 * b
    return (result_w, result_b)

# Create test data
rhs_w = [1.0, 2.0, 3.0, 4.0]
rhs_b = 0.5

ARGS = {
    "matvec": simple_matvec,
    "rhs_w": rhs_w,
    "rhs_b": rhs_b,
    "max_iter": 40,
    "tol": 1e-06,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
dot Ran
ran, violated the contract we drafted
BDS-SDU/utility-aware-data-pricing/data_pricing/valuation.py
code served (permissive licence) · get_code("9ac6af3c8dfd0a35")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Analyze the function:
# - left_w: Sequence[float] - used in zip with right_w
# - left_b: float - multiplied with right_b
# - right_w: Sequence[float] - used in zip with left_w
# - right_b: float - multiplied with left_b
# The function computes dot product of weights plus product of biases

# Create small sequences for weights
left_w = [1.0, 2.0, 3.0, 4.0]
right_w = [0.5, 1.5, 2.5, 3.5]

# Create scalar biases
left_b = 0.1
right_b = 0.2

ARGS = {
    "left_w": left_w,
    "left_b": left_b,
    "right_w": right_w,
    "right_b": right_b,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
sigmoid Ran
ran, honoured the contract we drafted
BDS-SDU/utility-aware-data-pricing/data_pricing/valuation.py
code served (permissive licence) · get_code("e431bb0c81b02097")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned float, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import math

def sigmoid(value: float) -> float:
    if value >= 0:
        z = math.exp(-value)
        return 1.0 / (1.0 + z)
    z = math.exp(value)
    return z / (1.0 + z)

ARGS = {
    "value": 0.5
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type float, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
simple_tokenize Ran
ran; our contract draft was wrong, not the code
BDS-SDU/utility-aware-data-pricing/data_pricing/valuation.py
code served (permissive licence) · get_code("45ce145302c2df1b")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import re

TOKEN_PATTERN = re.compile('[A-Za-z0-9_]+|[^\\w\\s]', re.UNICODE)

def simple_tokenize(text: str) -> list:
    return TOKEN_PATTERN.findall(text.lower())

# Construct valid arguments for simple_tokenize
ARGS = {
    "text": "Hello, World! This is a test_123 with special chars: @#$%"
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
stable_hash_index Ran
ran, honoured the contract we drafted
BDS-SDU/utility-aware-data-pricing/data_pricing/valuation.py
code served (permissive licence) · get_code("9b08c32610c7e0cf")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import hashlib

def stable_hash_index(token: str, dim: int) -> int:
    digest = hashlib.sha256(token.encode('utf-8')).hexdigest()
    return int(digest[:16], 16) % dim

ARGS = {
    "token": "hello_world",
    "dim": 256
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
HashedTextLogReg Not yet run
unverified (harvested, no recorded run)
BDS-SDU/utility-aware-data-pricing/data_pricing/valuation.py
code served (permissive licence) · get_code("a5c868738f88d82e")
How it ran

no invocation recorded.

influence_score Not yet run
unverified (harvested, no recorded run)
BDS-SDU/utility-aware-data-pricing/data_pricing/valuation.py
code served (permissive licence) · get_code("5b346c70d17bf3d8")
How it ran

no invocation recorded.

Also present in other papers' code

Of 8 samples lifted from this paper's linked repositories, 8 carry a function identity and 0 do not. Of the 8 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 8 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Traditional data valuation methods based on "row-count × quality coefficient" paradigms fail to capture the nuanced, nonlinear contributions that data makes to Large Language Model (LLM) capabilities. This paper presents a dynamic data valuation framework that transitions from static accounting to utility-based pricing. Our approach operates on three layers: (1) token-level information density metrics using Shannon entropy and Data Quality Scores; (2) empirical training gain measurement through influence functions, proxy model strategies, and Data Shapley values; and (3) cryptographic verifiability through hash-based commitments, Merkle trees, and a tamper-evident training ledger. We provide comprehensive experimental validation on three real domains (instruction following, mathematical reasoning, and code summarization), demonstrating that proxy-based empirical gain achieves near-perfect ranking alignment with realized utility, substantially outperforming row-count and token-count baselines. This framework enables a fair Data-as-a-Service economy where high-reasoning data is priced according to its actual contribution to model intelligence, while providing the transparency and auditability necessary for trustworthy data markets.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2604.22893")
get_code_for_paper("2604.22893")
have("2604.22893")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2604.22893.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2604.22893.svg)](https://syntology.ai/paper/2604.22893)

Connect an agent — have() is free.