AI Paper Atlas

Research map / Large language models / Alignment and preferences

FSPO: Few-Shot Optimization of Synthetic Preferences Effectively Personalizes to Real Users

Anikait Singh, Sheryl Hsu, Kyle Hsu, Eric Mitchell, Stefano Ermon, Tatsunori Hashimoto, Archit Sharma, Chelsea Finn

ICLR 2026 · Large language models › Alignment and preferences · approach: Reward model design and preference models

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add user description rationalization (RAT) to FSPO (ablation)FSPO · Reviews, trained userswinrate vs base (%): 80.4 -> 92.3Improved
add user description rationalization (RAT) to FSPO (ablation)FSPO · Roleplay (1500 users)winrate vs base (%): 82.6 -> 90.3Improved
meta-learn over users with FSPO instead of few-shot prompting aloneLlama 3.2 3B Instruct · Roleplay (1500 users)winrate vs base (%): 63.2 -> 82.6Improved
use IPO instead of DPO as the preference objective in FSPO (ablation)FSPO · Roleplay (1500 users)winrate vs base (%): 81.3 -> 82.6No clear effect
increase number of few-shot examples from 1 to 8 (ablation)FSPO · Roleplay held-out synthetic userswinrate vs base (%): 65.7 -> 82.6Improved
increase fraction of preference data per user from 10% to 100% (ablation)FSPO · Roleplay held-out synthetic userswinrate vs base (%): 70.1 -> 82.6Improved
increase preference data from 10% to 25% (ablation)FSPO · Roleplay held-out synthetic userswinrate vs base (%): 70.1 -> 69.5No clear effect
use persona steering instead of low-temperature sampling for response diversity (ablation)synthetic data generation · Roleplay response diversity (ALOE)mean semantic similarity: 0.96 -> 0.81Improved
use view steering plus ensemble of larger models instead of a single model with view steering (ablation)synthetic data generation · Roleplay response diversity (ALOE)mean semantic similarity: 0.78 -> 0.71Improved
iterative persona refinement of underspecified user descriptions (ablation)synthetic preference labeling · Roleplay preference labelsbinary Shannon entropy (nats): 0.64 -> 0.13Improved
FSPO versus base model with real human participantsFSPO · Roleplay human studyhuman winrate (%): 50 -> 68.2Improved
increase shots from 4 to 8 for few-shot prompting (ablation)Few-shot prompting baseline · Reviews, interpolated userswinrate vs base (%): 61.9 -> 59.1Hurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy