Home/LLM Frameworks/align-anything

align-anything

PKU-Alignment/align-anything

Training All-modality Model with Feedback

GitHub

GraphCanon updated Sep 20, 2026 · GitHub synced Sep 20, 2026

44views this month

4.7k stars505 forksLast push Nov 27, 2025 Python Apache-2.0

Decision brief

Align Anything focuses on training large models with multiple forms of feedback across various data modalities, leveraging RLHF and DPO.

Good fit when

- When you are developing a model that requires feedback from human evaluators and needs to handle different types of data (multimodal).
- If your project involves visual-textual or language-only large models requiring Reinforcement Learning from Human Feedback (RLHF) and Data Programming Optimization (DPO).

Avoid when

- When the model training does not benefit from advanced feedback mechanisms like RLHF or DPO.
- For projects that do not require support for multiple data modalities.

Requirements:: Python execution environment

Observed Jul 12, 2026 · Source: enrich:decision_facts

align-anything alternatives Trust report

Verify the decision

Maintenance and security

Full trust report

Maintenance: Slowing (263d since push); As of Aug 17, 2026
Provenance: Not a fork · Organization account; As of Aug 17, 2026
Security (OSV): No lockfile; As of Jul 11, 2026

Public GitHub metadata and optional OSV scans. Signals, not a guarantee. Trust methodology.

Install

pip install align-anything

PyPI

How it fits your stack(3)

Typed graph edges - alternatives, integrations, successors, and dependencies. Ranked by relationship type, not raw GitHub stars.

Integrates

LlamaFactory

Align-anything provides alignment methods such as Supervised Fine-Tuning (SFT), Direct Preference Optimization (DPO), Proximal Policy Optimization (PPO), and rule-based reinforcement learning to align multimodal large models with human intentions. LlamaFactory offers an efficient fine-tuning process for over 100 language models and vision-language models supported by major tech companies. The 'int

integrates

Awesome-Multimodal-Large-Language-Models

Both repositories focus on multimodal large language models, but 'Align Anything' focuses specifically on training techniques with feedback, while the former is a collection of resources and advances in the field.

MGM

Both 'Align Anything' and MGM deal with multimodal vision-language models, but they approach the problem differently with 'Align Anything' focusing on training methodologies and MGM on a framework for both understanding and generating images.

Relationship graph

Optional deeper exploration of typed edges and category neighbours.

Similar tools

Same-category neighbours not already linked as typed edges.

stanford_alpaca

Code and documentation to train Stanford's Alpaca models, and generate the data

30k★

trl

Train transformer language models with reinforcement learning

19k★

gorilla

Training and Evaluating LLMs for Function Calls (Tool Calls)

13k★

Evidence and technical details

Sourced facts, taxonomy, compatibility claims, README excerpt, and machine-readable endpoints.

Overview

Developed for training large, all-modality models using feedback mechanisms including Reinforcement Learning from Human Feedback (RLHF) and Data Programming Optimization (DPO). Supports a wide array of data types and modalities for alignment purposes.

Capability facts

Deploy: Self-host
Source: dockerfile:Dockerfile · Aug 17, 2026
Docker: Dockerfile present
Source: dockerfile:Dockerfile · Aug 17, 2026
Languages: python
Source: github.language+pyproject.toml · Aug 17, 2026

Graph entities

vLLMframework Smolagentsframework

README

License align anything is released under Apache License 2.0.

For agents

This page has a .md twin and JSON over the API.

/tools/pku-alignment-align-anything.md GET /api/graphcanon/tools/pku-alignment-align-anything GET /api/graphcanon/graph?tool=pku-alignment-align-anything