SuperAPI AI Directory
Partner with us中文AI newsFavorites

AI developer platforms

3,633 results · Page 30 of 101

simple_GRPO

AI developer platforms

simple_GRPO is a lightweight open-source GRPO implementation for reproducing the reasoning capabilities of R1-like large language models. The code is only about 200 lines, spread across two files, and depends on standard libraries such as deepspeed and torch, with no need for ray. It draws on the core loss calculation formula from Hugging Face trl, and supports saving GPU memory, quickly understanding the RL process, trying multi-answer generation, regrouping, KL penalty, and parameter tuning, among other things. An "Aha moment" can be observed in the early stages of training. The project provides a decoupled reference model design, allowing it to run on different GPUs or even different machines, and supports vLLM-accelerated inference and the reinforce++ algorithm. Experiments show that training is completed within 1 hour on an A800, and the Qwen2.5-7B and 3B models exhibit an Aha moment within 30 steps, with accuracy stably above 60% and format compliance reaching 100%.

Read more

simpleRL-reason

AI developer platforms

simpleRL-reason is an open-source reinforcement learning training codebase for enhancing reasoning ability, released on GitHub by the NLP team at the Hong Kong University of Science and Technology. It provides a simple reinforcement learning training scheme that can improve the reasoning ability of base models using only rule-based rewards and the GSM8K/Math datasets. The project has successfully conducted zero RL training on 10 different base models, including Llama3 8B, Mistral 7B/24B, DeepSeekMath 7B, Qwen2.5 0.5B/1.5B/7B/14B/32B, and Qwen2.5-Math-7B. Using limited data of 8K examples, the accuracy improvement reaches 10 to more than 20 absolute percentage points. The project has open-sourced the code, model checkpoints, and analysis tools, and provides related papers and blogs. The code is implemented based on Verl, supports custom environment settings and FSDP training, uses the GRPO algorithm, and accelerates training with Ray and vLLM.

Read more

Skillry

AI developer platforms

Skillry is a platform that provides reusable Agent Skills for AI agents such as Claude Code, Codex, and Cursor. Its skills are independently designed, written, and tested by the team, packaged in the form of SKILL.md, and cover tasks such as slides, web pages, images, and videos. Users can install skills via the web or CLI, enabling agents to generate corresponding finished products according to preset workflows. The platform has accumulated more than 400 skills, including free and paid content; paid offerings include monthly and annual plans, as well as a one-time lifetime purchase option. Some image or video skills require users to provide their own API keys for external AI models (such as FAL), with the corresponding provider charging based on usage. Free skills only require registering for a free account, and each skill page explains the actual output and required conditions.

Read more

skyagi

AI developer platforms

SkyAGI is a Python package for demonstrating the emergent abilities of large language models (LLMs) in simulating believable human behavior. It implements the idea of Generative Agents and provides a role-playing game that creates a very interesting user experience. Unlike previous AI-based NPC systems, SkyAGI's NPCs can generate very believable human reactions. The interesting observations in this demo show that there is great potential for rethinking many aspects of game development, such as NPC scripting. To demonstrate this, SkyAGI provides example characters from The Big Bang Theory and The Avengers as a starting point. Users can also define custom characters by creating a configuration file similar to customized_character.json.

Read more

Skywork-OR1

AI developer platforms

Skywork-OR1 (Open Reasoner 1) is a series of math and code reasoning models and accompanying training solutions open-sourced by SkyworkAI. This series uses large-scale rule-based reinforcement learning, combined with carefully designed datasets and training recipes, and includes two general reasoning models, Skywork-OR1-7B and Skywork-OR1-32B. According to official materials, Skywork-OR1-32B outperforms Deepseek-R1 and Qwen3-32B on math tasks (AIME24 and AIME25) and performs comparably on code tasks (LiveCodeBench); Skywork-OR1-7B is competitive in math and coding scenarios among models of the same scale. The project has open-sourced model weights, training data, and code, and provides training and evaluation scripts, supports reproduction via Docker and Conda environments, and supports multi-node training using Ray. The official team also released a technical report (arXiv) and Notion Blog, sharing detailed training recipes, experimental results, and analysis. The project also provides the Skywork-OR1-RL-Data training dataset, as well as evaluation scripts for AIME24/AIME25 and LiveCodeBench. Evaluation uses Avg@K as the primary metric rather than Pass@1, to reduce the impact of randomness.

Read more

slime

AI developer platforms

slime is an LLM post-training framework developed by THUDM, designed for scaled reinforcement learning (RL) training. By connecting Megatron with SGLang, it provides two core capabilities: high-performance training and flexible data generation, and unifies training, rollout, data buffering, reward computation, verifier feedback, and environment interaction into a single path. The framework has been used for the release training of models such as GLM-5.3-Flash, GLM-5.3, GLM-5.2, GLM-5.1, GLM-5, GLM-4.7, GLM-4.6, and GLM-4.5, while also supporting model families such as Qwen, DeepSeek, and Llama 3. The design goal of slime is to let training and data generation reinforce each other, without having to cobble together a heavy tech stack composed of fragmented trainers, rollout services, and agent frameworks. As a result, it is small in size, easy to understand and extend, and yet validated through a complete training loop.

Read more

Slow_Thinking_with_LLMs

AI developer platforms

STILL (Slow Thinking with LLMs) is a series of technical research projects on slow-thinking reasoning systems for large language models conducted by the RUCAIBox team at Renmin University of China, with long-term exploration of o1-like reasoning systems. The project has open-sourced a large amount of code, models, training data, and evaluation benchmarks, covering directions such as reinforcement learning exploration mechanisms, tool-augmented reasoning, code-integrated reasoning, autonomous web search, and hallucination mitigation, and has released STILL-1, STILL-2, STILL-3, and multiple derivative models and benchmarks.

Read more

smol-vision

AI developer platforms

smol-vision is an open-source collection of recipes for vision and multimodal AI models, with the goal of shrinking, optimizing, and customizing frontier models. The repository provides several runnable examples, including fine-tuning Kosmos2.5 on OCR and bounding box tasks, fine-tuning Florence-2 on document question answering tasks, and fine-tuning DINOv3 on image classification tasks. The repository is maintained by merveenoyan and is published as a GitHub repository. Because GitHub has been unable to render notebooks properly for a long time, notebooks containing rich outputs have been moved to another location, but the repository is still being continuously updated. The project's specific dependencies, installation commands, license agreement, and ways to participate are not yet provided.

Read more