RLHFlow/Directional-Preference-Alignment
Directional Preference Alignment
This project helps you fine-tune large language models (LLMs) to generate responses that match specific user preferences, like being more helpful or more verbose. You input a user prompt and specify the desired 'mix' of attributes (e.g., 70% helpfulness, 30% verbosity), and the model outputs a text response tailored to those instructions. This is ideal for content creators, customer service managers, or anyone needing precise control over an LLM's output style and content.
No commits in the last 6 months.
Use this if you need to precisely control the stylistic and content attributes of an LLM's generated text, beyond simple prompts.
Not ideal if you're looking for a general-purpose LLM without needing fine-grained control over specific output characteristics.
Stars
58
Forks
4
Language
—
License
Apache-2.0
Category
Last pushed
Sep 23, 2024
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/llm-tools/RLHFlow/Directional-Preference-Alignment"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
codelion/pts
Pivotal Token Search
DtYXs/Pre-DPO
Pre-DPO: Improving Data Utilization in Direct Preference Optimization Using a Guiding Reference Model
dannylee1020/openpo
Building synthetic data for preference tuning
pspdada/Uni-DPO
[ICLR 2026] Official repository of "Uni-DPO: A Unified Paradigm for Dynamic Preference...
liushunyu/awesome-direct-preference-optimization
A Survey of Direct Preference Optimization (DPO)