SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2311.13628 · NeurIPS · 2023

Prompt Risk Control: A Rigorous Framework for Responsible Deployment of Large Language Models

Richard Zemel, Jake Snell, Zhun Deng, Thomas Zollo, Todd Morrill

arXiv · PDF · Open in the Atlas

Code that ran

2 of 5 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 2 failures of our instrument; 3 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/2311.13628.json · badge: /api/ran/2311.13628.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
thomaspzollo/prompt_risk canonical (classifier label)
classifier basis: classifier_2025
2 ran · 3 no recorded run
FunctionStatusWhere it lives
find_rcps_ucb Ran
ran; our contract draft was wrong, not the code
thomaspzollo/prompt_risk/prompt_risk/methods.py
code served (permissive licence) · get_code("b966af02dc34fd88")
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, through the batch API) and returned Tensor, ndim 1, dtype family float, shape (3,) 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
from torch import Tensor
from typing import Callable

torch.manual_seed(0)

# Create a small loss tensor with shape (H, N) where H=3 hypotheses, N=4 examples
loss = torch.randn(3, 4)

# Create a simple UCB function that takes a 1D tensor and delta, returns a float
def ucb_fun(x: Tensor, delta: float) -> float:
    """Simple UCB function that computes mean + sqrt(log(1/delta) / (2*n))"""
    n = x.shape[0]
    if n == 0:
        return float('inf')
    mean_loss = x.mean().item()
    confidence_radius = (torch.log(torch.tensor(1.0 / delta)) / (2.0 * n)).sqrt().item()
    return mean_loss + confidence_radius

# Set delta and target parameters
delta = 0.1
target = 0.5

ARGS = {
    "loss": loss,
    "delta": delta,
    "target": target,
    "ucb_fun": ucb_fun
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 1, dtype family float, shape (3,) — shape (3,) 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
wsr_ucb Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
thomaspzollo/prompt_risk/prompt_risk/methods.py
code served (permissive licence) · get_code("17d4d18738b76ea1")
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, through the batch API) 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)

# Create a small tensor of loss values in range [0, 1]
loss = torch.tensor([0.1, 0.3, 0.2, 0.5, 0.4], dtype=torch.float32)

# Create a small delta value for confidence bound
delta = 0.05

ARGS = {
    "loss": loss,
    "delta": delta,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy, 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
RcpsWSR Not yet run
unverified (harvested, no recorded run)
thomaspzollo/prompt_risk/prompt_risk/methods.py
code served (permissive licence) · get_code("c028f3bc72993dc9")
How it ran

no invocation recorded.

WSR_mu_plus Not yet run
unverified (harvested, no recorded run)
thomaspzollo/prompt_risk/prompt_risk/methods.py
code served (permissive licence) · get_code("0b988a3c5dbab48a")
How it ran

no invocation recorded.

select_k_shot_examples Not yet run
unverified (harvested, no recorded run)
thomaspzollo/prompt_risk/scripts/generate_outputs.py
code served (permissive licence) · get_code("74b7aa2a6115ee37")
How it ran

no invocation recorded. this sample carries JSON test cases instead of a setup block; get_reference_implementation serves them with the level they back.

Also present in other papers' code

Of 5 samples lifted from this paper's linked repositories, 5 carry a function identity and 0 do not. Of the 5 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 0 later) and 4 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.

FunctionAlso present inEarliest-datedOthers named
WSR_mu_plus
brick of this paper's assembly: helper
1 other paper
1 dated earlier
2212.13629 (2022-12)
Quantile Risk Control: A Flexible Framework for Bounding the Probability of High-Loss Predictions

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

The recent explosion in the capabilities of large language models has led to a wave of interest in how best to prompt a model to perform a given task. While it may be tempting to simply choose a prompt based on average performance on a validation set, this can lead to a deployment where unexpectedly poor responses are generated, especially for the worst-off users. To mitigate this prospect, we propose Prompt Risk Control, a lightweight framework for selecting a prompt based on rigorous upper bounds on families of informative risk measures. We offer methods for producing bounds on a diverse set of metrics, including quantities that measure worst-case responses and disparities in generation quality across the population of users. In addition, we extend the underlying statistical bounding techniques to accommodate the possibility of distribution shifts in deployment. Experiments on applications such as open-ended chat, medical question summarization, and code generation highlight how such a framework can foster responsible deployment by reducing the risk of the worst outcomes. Recent leaps in the capabilities of large language models (LLMs) such as GPT-4 (OpenAI, 2023), LLaMA (Touvron et al., 2023), and Claude have driven a wave of interest in constructing the best prompt for a given task, where a prompt generally refers to an input to the LLM. Various prompting strategies have been proposed, including but not limited to: in-context learning (Brown et al., 2020), instruction following (Wei et al., 2022a), chain-of-thought prompting (Wei et al., 2022b), Published as a conference paper at ICLR 2024. * indicates equal contribution.

For agents

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

get_harvested_code_for_paper("2311.13628")
get_code_for_paper("2311.13628")
have("2311.13628")

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

curl https://syntology.ai/api/ran/2311.13628.json

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

[![Syntology run record](https://syntology.ai/api/ran/2311.13628.svg)](https://syntology.ai/paper/2311.13628)

Connect an agent — have() is free.