Research map / Large language models / AI and society
TRIPLE: Theory-Driven Integration of Planned and Habitual Behaviors for LLM-based Personalization
Taehyung Noh, Seungwan Jin, Haein Yeo, Kyungsik Han
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add behavioral rationale to habit+intention profiles (ablation) | TRIPLE (Llama-3.1-70B) · LaMP-5 | BLEU: 10.484 -> 11.467 | Improved |
| add behavioral rationale to habit+intention profiles (ablation) | TRIPLE (Llama-3.1-70B) · LaMP-4 | BLEU: 3.810 -> 4.994 | Improved |
| add behavioral rationale to habit+intention profiles (ablation) | TRIPLE (Llama-3.1-70B) · LaMP-3 | MAE: 0.290 -> 0.280 | Improved |
| add behavioral rationale to habit+intention profiles (ablation) | TRIPLE (Llama-3.1-70B) · LaMP-2M | accuracy: 0.648 -> 0.610 | Hurt |
| add intention profile and rationale on top of habit profile (ablation) | TRIPLE (Llama-3.1-70B) · LaMP-2N | accuracy: 0.805 -> 0.795 | Hurt |
| add intention profile to habit profile, without rationale (ablation) | TRIPLE (Llama-3.1-70B) · LaMP-4 | BLEU: 4.394 -> 3.810 | Hurt |
| add habit profile to raw history (ablation) | In-context baseline (Llama-3.1-70B) · LaMP-5 | BLEU: 4.048 -> 10.178 | Improved |
| use intention profile only instead of habit profile (ablation) | TRIPLE (Llama-3.1-70B) · LaMP-5 | BLEU: 10.178 -> 9.049 | Hurt |
| TRIPLE vs best baseline on generation | TRIPLE (Llama-3.1-70B) · LaMP-5 | BLEU: 7.341 -> 11.467 | Improved |
| TRIPLE vs best baseline on generation | TRIPLE (Llama-3.1-8B) · LaMP-4 | BLEU: 0.961 -> 3.868 | Improved |
| TRIPLE vs best baseline on classification | TRIPLE (Llama-3.1-8B) · LaMP-2N | accuracy: 0.805 -> 0.800 | No clear effect |
| TRIPLE vs PURE on scholarly title ROUGE-1 | TRIPLE (Llama-3.1-70B) · LaMP-5 | ROUGE-1: 0.511 -> 0.493 | Hurt |
Closest papers by meaning
- HYDRA: Model Factorization Framework for Black-Box LLM Personalization (NeurIPS 2024 · similarity 0.86)
- How Does Personalized Memory Shape LLM Behavior? Benchmarking Rational Preference Utilization in Personalized Assistants (NeurIPS 2026 · similarity 0.86)
- Rethinking Psychometric Evaluation of LLMs: When and Why Self-Reports Predict Behavior (NeurIPS 2026 · similarity 0.86)
- PersonalLLM: Tailoring LLMs to Individual Preferences (ICLR 2025 · similarity 0.85)
- BehaviorBench: Modeling Real-World User Decisions from Behavioral Traces (NeurIPS 2026 · similarity 0.84)
- PANTHER: Generative Pretraining Beyond Language for Sequential User Behavior Modeling (NeurIPS 2025 · similarity 0.84)
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.
