SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2501.18532 · ICLR · 2025

PSA: DIFFERENTIALLY PRIVATE STEERING FOR LARGE LANGUAGE MODEL ALIGNMENT

Iryna Gurevych, Amartya Sanyal, Anmol Goel, Yaxi Hu

arXiv · PDF · Open in the Atlas

Code that ran

1 of 1 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; 1 failure of our instrument; 0 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; 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/2501.18532.json · badge: /api/ran/2501.18532.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
ukplab/iclr2025-psa canonical (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
FunctionStatusWhere it lives
make_pos_neg_pair Ran
ran; our contract draft was wrong, not the code
ukplab/iclr2025-psa/iclr2025_psa/utils.py
code served (permissive licence) · get_code("48589efa437d8d18")
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, 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
from transformers import AutoTokenizer

# Create a mock tokenizer class for Llama
class MockLlamaTokenizer:
    def __init__(self):
        self.name_or_path = "meta-llama/Llama-2-7b"

# Create a mock tokenizer class for Qwen
class MockQwenTokenizer:
    def __init__(self):
        self.name_or_path = "Qwen/Qwen-7B"
    
    def apply_chat_template(self, messages, tokenize=False, add_generation_prompt=False):
        # Simple mock implementation
        result = ""
        for msg in messages:
            result += f"<|im_start|>{msg['role']}\n{msg['content']}<|im_end|>\n"
        return result

# Create a mock tokenizer class for Gemma
class MockGemmaTokenizer:
    def __init__(self):
        self.name_or_path = "google/gemma-7b"
    
    def apply_chat_template(self, messages, tokenize=False, add_generation_prompt=False):
        # Simple mock implementation
        result = ""
        for msg in messages:
            result += f"<start_of_turn>{msg['role']}\n{msg['content']}<end_of_turn>\n"
        return result

# Create sample MWE data
mwe_data = {
    'question': 'What is the capital of France?',
    'answer_matching_behavior': 'Paris is the capital of France.',
    'answer_not_matching_behavior': 'London is the capital of France.'
}

# Create tokenizer instance (using Llama for this example)
tokenizer = MockLlamaTokenizer()

# Define ARGS
ARGS = {
    'mwe_data': mwe_data,
    'tokenizer': tokenizer
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch, transformers — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple 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

Also present in other papers' code

Of 1 samples lifted from this paper's linked repositories, 1 carry a function identity and 0 do not. Of the 1 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 1 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

Aligning Large Language Models (LLMs) with human values and away from undesirable behaviors (such as hallucination) has become increasingly important. Recently, steering LLMs towards a desired behavior via activation editing has emerged as an effective method to mitigate harmful generations at inference-time. Activation editing modifies LLM representations by preserving information from positive demonstrations (e.g., truthful) and minimising information from negative demonstrations (e.g., hallucinations). When these demonstrations come from a private dataset, the aligned LLM may leak private information contained in those private samples. In this work, we present the first study of aligning LLM behavior with private datasets. Our work proposes the Private Steering for LLM Alignment (PSA) algorithm to edit LLM activations with differential privacy (DP) guarantees. We conduct extensive experiments on seven different benchmarks with opensource LLMs of different sizes (0.5B to 7B) and model families (LlaMa, Qwen, Mistral and Gemma). Our results show that PSA achieves DP guarantees for LLM alignment with minimal loss in performance, including alignment metrics, openended text generation quality, and general-purpose reasoning. We also develop the first Membership Inference Attack (MIA) for evaluating and auditing the empirical privacy for the problem of LLM steering via activation editing. Our experiments support the theoretical guarantees by showing improved guarantees for our PSA algorithm compared to several existing non-private techniques. 1

For agents

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

get_harvested_code_for_paper("2501.18532")
get_code_for_paper("2501.18532")
have("2501.18532")

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

curl https://syntology.ai/api/ran/2501.18532.json

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

[![Syntology run record](https://syntology.ai/api/ran/2501.18532.svg)](https://syntology.ai/paper/2501.18532)

Connect an agent — have() is free.