Analyzing the Flow of Developer Prompts in ChatGPT Conversations
12 Nov 2025
How developers interact with ChatGPT across multiple turns—analyzing prompts, feedback, and flow patterns from 645 developer conversations.
What Do Developers Ask ChatGPT the Most?
12 Nov 2025
Discover what 580 GitHub conversations reveal about how developers use ChatGPT — from code generation to debugging and documentation.
Building the DevGPT Dataset for Developer–ChatGPT Studies
12 Nov 2025
How researchers collected and cleaned 17K developer–ChatGPT conversations from GitHub to explore AI’s role in software development.
Lessons on Developer–AI Collaboration From 580 GitHub Conversations
12 Nov 2025
Developers are using ChatGPT to code, debug, and collaborate. A new study reveals how shared AI chats are reshaping teamwork on GitHub.
Comparing Efficiency Strategies for LLM Deployment and Summarizing PowerInfer‑2’s Impact
3 Nov 2025
This article situates PowerInfer‑2 among other frameworks that improve LLM efficiency through compression, pruning, and speculative decoding.
Performance Evaluation of PowerInfer‑2: Offloading, Prefill, and In‑Memory Efficiency
3 Nov 2025
PowerInfer‑2 achieves up to 29× speedups over llama.cpp and 13× over LLMFlash by leveraging neuron‑level pipelines and NPU‑centric prefill optimization.
How PowerInfer‑2 Turns Your Smartphone Into an AI Workstation
3 Nov 2025
The cost model leverages SMT‑based solving (Z3) to achieve optimal decoding speed under CPU, I/O, and memory constraints.
How Hybrid AI Models Balance Memory and Efficiency
28 Oct 2025
Mamba and attention are combined in SAMBA to provide effective long-context language modeling with robust performance and exceptional memory recall.
Meet SAMBA: The AI Model That Remembers More and Trains Faster
28 Oct 2025
SAMBA demonstrates how combining recurrence and attention allows for quicker, longer, and more intelligent AI with infinite memory and superior efficiency.