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.

(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.

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

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

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 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 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

Analysis of emergent behavior in real human–AI dialogues.

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.

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

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

📊 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.