(ACL 2026 Main) LLMSurgeon recovers the pretraining data mixture of any LLM from only its generated text — no weights, no training data. A calibrated domain classifier plus label-shift correction de-blurs biased predictions. Ships with LLMScan, a benchmark on 8 open-source LLMs.

Quantifying prompt quality using information theory: entropy and mutual information analysis of 1,800 LLM generations

An RL environment where an LLM agent learns to curate talking-head video clips for AV LoRA training. No labels exposed, rewards only.

Comparative tool for critical code studies and other methods for comparative analysis of LLM output.

Stock Analysis Dashboard featuring Risk, Fundamental, Sentiment, and Technical analysis, plus AI-powered insights with a rating score, summary table, overall evaluation, and detailed breakdown of each analysis type.

lawhead-extractor parses legal headlines, extracting parties, claim type and outcome using an LLM with pattern matching for accuracy.

A new package that takes user-provided text (such as a blog post title or a short article snippet) and generates a structured summary highlighting key advantages or claims. It uses an LLM to analyze t

Real-time supply chain monitoring for Python and NPM ecosystems with LLM/AI-powered diff analysis.

Analysis of emergent behavior in real human–AI dialogues.

A new package would process user complaints or descriptions about logging systems, extracting structured insights such as common pain points, root causes, or improvement suggestions. It uses an LLM to

A Python CLI tool that collects and analyzes Discourse forum discussions using Claude AI to identify common problems, categorize issues by severity, and provide natural language querying of forum insights.

A Python-based tool for comparing translated .docx documents against their original versions. It highlights differences, calculates similarity metrics, and generates detailed comparison reports, including suggested corrections.

AI Text Slop: A Quantitative Study of Stylistic Convergence Across Six Language Models in Japanese Technical Writing

A framework for analyzing Large Language Model (LLM) performance through Quantized PSA and structured weight pruning experiments

📊 Explore how Shannon entropy and mutual information can quantify prompt quality in generative AI systems across various temperature settings.

🔍 Analyze user feedback on logging systems with Logference, extracting insights to identify pain points and improve efficiency.