Papers › DSPA: Dynamic SAE Steering for Data-Efficient Preference Alignment

DSPA: Dynamic SAE Steering for Data-Efficient Preference Alignment

23 Mar 2026arXiv:2603.21461added by Syntology

James Wedgwood, Aashiq Muhamed, Mona T. Diab, Virginia Smith

Title, abstract, authors and date from arXiv's metadata (CC0); this paper is not in the Papers with Code archive (frozen 2025-07-28).

Preference alignment is usually achieved by weight-updating training on preference data, which adds substantial alignment-stage compute and provides limited mechanistic visibility. We propose Dynamic SAE Steering for Preference Alignment (DSPA), an inference-time method that makes sparse autoencoder (SAE) steering prompt-conditional. From preference triples, DSPA computes a conditional-difference map linking prompt features to generation-control features; during decoding, it modifies only token-active latents, without base-model weight updates. Across Gemma-2-2B/9B and Qwen3-8B, DSPA improves MT-Bench and is competitive on AlpacaEval while preserving multiple-choice accuracy. Under restricted preference data, DSPA remains robust and can rival the two-stage RAHF-SCIT pipeline while requiring up to 4.47× fewer alignment-stage FLOPs. Finally, we audit the SAE features DSPA modifies, finding that preference directions are dominated by discourse and stylistic signals, and provide theory clarifying the conditional-difference map estimate and when top-k ablation is principled.

PaperPDFCode Syntology ran

In Syntology Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, Syntology's MCP tool lists every function and class Syntology harvested from this paper and whether it ran (how to connect): get_harvested_code_for_paper(arxiv_id="2603.21461")

Code

Syntology Ran 1 of 7 code samples harvested from 1 repository linked to this paper; 6 have no recorded run. Of those that ran: 1 ran · violated contract.

By repository: found in paper text by Syntology: 7 samples from 1 repository, 1 ran. The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

tatsu-lab/alpaca_eval found in paper text by Syntology report
lm-sys/FastChat found in paper text by Syntology report

Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

7 samples harvested; 1 ran; 0 honoured the contract we drafted; 6 have no recorded run. Read from Syntology's graph 2026-09-24; that is when this build read the record, not when the samples ran.

1ran · violated contract
6unverified

Licence: 0 of the 7 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from lm-sys/FastChat. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself; “our draft was wrong” and “fixture could not drive it” are failures of Syntology's instrument, not of the code.

Each sample ends with its code_sha256, Syntology's identity for that exact code. An agent fetches the stored sample with Syntology's MCP tool get_code(code_sha256="…") (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples from a repository marked official are listed first. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of four reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; or the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar). Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

raise_warning_for_incompatible_cpu_offloading_configuration lm-sys/FastChat/fastchat/model/model_adapter.py found in paper text by Syntology ran · violated contract Apache-2.0 (permissive) · 0bf414a2b44babe6 · report
compress lm-sys/FastChat/fastchat/model/compression.py found in paper text by Syntology unverified Apache-2.0 (permissive) · 80dde39da1ba3708 · report
get_compressed_list lm-sys/FastChat/fastchat/model/compression.py found in paper text by Syntology unverified Apache-2.0 (permissive) · c5824da57107fe8c · report
get_conv_template lm-sys/FastChat/fastchat/conversation.py found in paper text by Syntology unverified Apache-2.0 (permissive) · 0e29f2d151ce2bd7 · report
get_jacobian_trajectory lm-sys/FastChat/fastchat/model/model_cllm.py found in paper text by Syntology unverified Apache-2.0 (permissive) · a42fafd21685fb54 · report
process_response lm-sys/FastChat/fastchat/model/model_chatglm.py found in paper text by Syntology unverified Apache-2.0 (permissive) · dc21db7cf25bfe72 · report
recover_message_list lm-sys/FastChat/fastchat/model/model_chatglm.py found in paper text by Syntology unverified Apache-2.0 (permissive) · e4f8545686f50507 · report

Results from the paper

The Papers with Code archive ends with its 2025-07-28 snapshot. This paper's arXiv identifier, 2603.21461, was issued in March 2026, after that date, so the archive has no leaderboard rows for it.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections