Article
We Got Claude to Fine-Tune an Open Source LLM
•
558
Aayush
Aayushfaced
AI & ML interests
None yet
Recent Activity
upvoted
an
article
18 days ago
We Got Claude to Fine-Tune an Open Source LLM
upvoted
an
article
18 days ago
New in llama.cpp: Model Management
upvoted
a
collection
about 1 month ago
NVIDIA Nemotron V2
Organizations
None yet
upvoted
2
articles
18 days ago
Article
New in llama.cpp: Model Management
•
103
upvoted
2
collections
about 1 month ago
upvoted
an
article
about 1 month ago
Article
🤔👀🎬🖥️📖 Kimi-VL-A3B-Thinking-2506: A Quick Navigation
•
74
liked
a
Space
2 months ago
liked
3
Spaces
3 months ago
upvoted
2
papers
3 months ago
liked
a
dataset
4 months ago
upvoted
7
papers
4 months ago
FastVLM: Efficient Vision Encoding for Vision Language Models
Paper
•
2412.13303
•
Published
•
72
VGGT: Visual Geometry Grounded Transformer
Paper
•
2503.11651
•
Published
•
35
FlowRL: Matching Reward Distributions for LLM Reasoning
Paper
•
2509.15207
•
Published
•
114
AgentGym-RL: Training LLM Agents for Long-Horizon Decision Making through Multi-Turn Reinforcement Learning
Paper
•
2509.08755
•
Published
•
56
A Survey of Reinforcement Learning for Large Reasoning Models
Paper
•
2509.08827
•
Published
•
190
LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models
Paper
•
2403.13372
•
Published
•
175
A Comprehensive Survey of Self-Evolving AI Agents: A New Paradigm Bridging Foundation Models and Lifelong Agentic Systems
Paper
•
2508.07407
•
Published
•
98
reacted to
Kseniase's
post with 👍
4 months ago
Post
7052
10 Latest Preference Optimization Techniques
Models need feedback on what makes outputs “good” or “bad.” Policy optimization (PO) turns preferences and rewards into actual training signals. This field is evolving quickly, moving far beyond classics like PPO and GRPO. So here is our overview of 10 newest PO methods:
1. Pref-GRPO → Pref-GRPO: Pairwise Preference Reward-based GRPO for Stable Text-to-Image Reinforcement Learning (2508.20751)
Stabilizes text-to-image reinforcement learning (RL) with pairwise preference rewards and a unified UNIGENBENCH benchmark
2. PVPO (Policy with Value Preference Optimization) → PVPO: Pre-Estimated Value-Based Policy Optimization for Agentic Reasoning (2508.21104)
This critic-free RL method uses a pre-trained model as a reference anchor to reduce bias and guide learning, selecting high-value examples through data pre-sampling
3. DCPO (Dynamic Clipping Policy Optimization) → DCPO: Dynamic Clipping Policy Optimization (2509.02333)
Uses dynamic clipping, which adjusts probability limits per token for better token exploration, and smooth reward standardization to balance rewards over training steps and prevent wasted updates
4. ARPO (Agentic Reinforced Policy Optimization) → Agentic Reinforced Policy Optimization (2507.19849)
Optimizes multi-turn LLM agents that use external tools. It uses an entropy-based adaptive rollout to explore post-tool use and an advantage attribution method to better assign credit across steps, leading to more efficient tool use with fewer resources
5. GRPO-RoC (Group Relative Policy Optimization with Resampling-on-Correct) → rStar2-Agent: Agentic Reasoning Technical Report (2508.20722)
Oversamples rollouts, then resamples them to keep diverse mistakes and only the highest-quality correct answers. It reduces noises and ends up with stronger reasoning in a code environment
Read further below ⬇️
If you like this, also subscribe to the Turing post: https://www.turingpost.com/subscribe
Models need feedback on what makes outputs “good” or “bad.” Policy optimization (PO) turns preferences and rewards into actual training signals. This field is evolving quickly, moving far beyond classics like PPO and GRPO. So here is our overview of 10 newest PO methods:
1. Pref-GRPO → Pref-GRPO: Pairwise Preference Reward-based GRPO for Stable Text-to-Image Reinforcement Learning (2508.20751)
Stabilizes text-to-image reinforcement learning (RL) with pairwise preference rewards and a unified UNIGENBENCH benchmark
2. PVPO (Policy with Value Preference Optimization) → PVPO: Pre-Estimated Value-Based Policy Optimization for Agentic Reasoning (2508.21104)
This critic-free RL method uses a pre-trained model as a reference anchor to reduce bias and guide learning, selecting high-value examples through data pre-sampling
3. DCPO (Dynamic Clipping Policy Optimization) → DCPO: Dynamic Clipping Policy Optimization (2509.02333)
Uses dynamic clipping, which adjusts probability limits per token for better token exploration, and smooth reward standardization to balance rewards over training steps and prevent wasted updates
4. ARPO (Agentic Reinforced Policy Optimization) → Agentic Reinforced Policy Optimization (2507.19849)
Optimizes multi-turn LLM agents that use external tools. It uses an entropy-based adaptive rollout to explore post-tool use and an advantage attribution method to better assign credit across steps, leading to more efficient tool use with fewer resources
5. GRPO-RoC (Group Relative Policy Optimization with Resampling-on-Correct) → rStar2-Agent: Agentic Reasoning Technical Report (2508.20722)
Oversamples rollouts, then resamples them to keep diverse mistakes and only the highest-quality correct answers. It reduces noises and ends up with stronger reasoning in a code environment
Read further below ⬇️
If you like this, also subscribe to the Turing post: https://www.turingpost.com/subscribe
- 1 reply