ICLR2025-Papers-with-Code

🥈Silver

Access and automate research from ICLR 2021-2025 papers and open-source projects. Ideal for operations teams to stay updated on LLM advancements in NLP. Integrates with Claude for automated research extraction and analysis.

574280Updated 1mo ago

Intermediate30min to implementautomation

Saves ~20 min per use

Quick InstallView Source

git clone https://github.com/yinizhilian/ICLR2025-Papers-with-Code.git

Works with:

Claude

Overview

About This Skill

How to Use

[{"step":"Identify your focus area by replacing [TOPIC] in the prompt template with a specific research domain (e.g., 'vision transformers' or 'multimodal learning').","tip":"Use broad terms like 'LLM efficiency' or 'multimodal models' to cast a wide net, then refine based on initial results."},{"step":"Run the prompt in your AI assistant (e.g., Claude, ChatGPT) and review the generated summary for relevance.","tip":"Check the 'Why It Matters' section to quickly assess the practical impact of each paper."},{"step":"For each paper of interest, visit the linked repositories to explore implementation details, benchmarks, and community discussions.","tip":"Look for README files, issue trackers, and example notebooks to gauge ease of adoption."},{"step":"Use the summaries to inform your team's roadmap or to generate a follow-up report on implementation feasibility.","tip":"Highlight papers with active maintenance (recent commits, responsive authors) for higher adoption potential."},{"step":"Set up automated alerts (e.g., via Google Scholar or Papers with Code) for new ICLR 2025 papers matching your topic.","tip":"Use tools like [Papers with Code](https://paperswithcode.com/) to filter by 'ICLR 2025' and your topic, then subscribe to RSS feeds."}]

Use Cases

Automate the retrieval of the latest ICLR papers and their corresponding code implementations for research purposes.

Facilitate the integration of state-of-the-art NLP techniques into existing projects by leveraging ICLR resources.

Streamline the process of comparing different models and methodologies presented in ICLR papers.

Enhance educational resources by providing students with access to the latest research and practical code examples.

Setup & Installation

Quick Install

No install command available. Check the GitHub repository for manual installation instructions.

Alternative Install (Git Clone)

git clone https://github.com/yinizhilian/ICLR2025-Papers-with-Code

Requirements

Claude Code or compatible AI agent
Works with: Claude

Quick Start Guide

Install the Skill

Copy the install command above and run it in your terminal.

Open Your AI Agent

Launch Claude Code, Cursor, or your preferred AI coding agent.

Try It Out

Use the prompt template or examples below to test the skill.

Customize

Adapt the skill to your specific use case and workflow.

Usage Examples

Prompt Template

Generate a concise research summary of the latest ICLR 2025 papers focused on [TOPIC, e.g., 'efficient fine-tuning of LLMs']. Include key contributions, implementation details, and open-source code links. Prioritize papers with available repositories. Format as: 1) Paper Title (Authors), 2) Key Idea (1-2 sentences), 3) Code/Repo (if available), 4) Why It Matters (1 sentence).

Example Output

### ICLR 2025 Research Summary: Efficient Fine-Tuning of LLMs

1) **LoRA-FT: Low-Rank Adaptation for Parameter-Efficient Fine-Tuning** (Smith et al.)
   - Introduces a low-rank decomposition method to reduce fine-tuning memory usage by 80% compared to full fine-tuning, while maintaining 95% of original model performance.
   - Implemented in PyTorch; code available at: [github.com/smith/lora-ft](https://github.com/smith/lora-ft)
   - Enables fine-tuning of 70B parameter models on a single A100 GPU, democratizing access to large-scale LLM adaptation.

2) **FlashAttention-2: Faster and More Memory-Efficient Attention for LLMs** (Dao et al.)
   - Optimizes attention computation by reducing memory reads/writes by 3x and improving throughput by 25% on modern GPUs.
   - Code integrated into Hugging Face Transformers; repo: [github.com/huggingface/transformers](https://github.com/huggingface/transformers)
   - Critical for scaling inference and training of LLMs without hardware upgrades.

3) **QLoRA: Quantized Low-Rank Adaptation for 4-bit LLMs** (Dettmers et al.)
   - Combines 4-bit quantization with LoRA to enable fine-tuning of 65B parameter models on a single RTX 3090 (12GB VRAM).
   - Code: [github.com/artidoro/qlora](https://github.com/artidoro/qlora)
   - Reduces hardware barriers for researchers and practitioners working with state-of-the-art LLMs.

**Why This Matters**: These papers collectively lower the barrier to entry for fine-tuning and deploying LLMs, making advanced NLP accessible to teams with limited compute resources.