Planned Work & Future Publications

This page outlines ongoing research projects and planned publications representing the next phase of my research in cross-modal learning, optimization, and human-AI collaboration.

Current Manuscripts in Development

2025 Target Submissions

“Secret Hitler AI: Reinforcement Learning for Strategic Deception in Social Games”

Authors: Stephen Mander, et al.
Target Journal: AI Magazine (Special Issue on Game AI)
Status: 70% complete, expected submission March 2025

Abstract: This paper presents a novel application of reinforcement learning to the social deduction game Secret Hitler, exploring AI strategies for deception, coalition building, and information gathering in multi-agent environments with incomplete information.

Key Innovations:

  • LLM-based agent with LoRA fine-tuning for strategic reasoning

  • Novel evaluation framework for social deduction AI performance

  • Analysis of strategic decision-making in incomplete information games

  • Open-source framework for social game AI research

Current Progress:

  • ✅ Base LLM implementation and training pipeline

  • ✅ Initial RL training experiments

  • 🔄 Strategic evaluation framework development

  • ⏳ Comprehensive multi-agent tournament analysis


“MCMC-Enhanced LSA: Reusable Kernels for Cross-Domain Semantic Analysis”

Authors: Stephen Mander, et al.
Target Journal: Computational Linguistics
Status: 60% complete, expected submission April 2025

Abstract: We present novel Markov Chain Monte Carlo approaches for Latent Semantic Analysis that create reusable computational kernels, enabling more robust semantic models with improved transferability across different text domains and languages.

Key Innovations:

  • MCMC sampling techniques for LSA parameter estimation

  • Reusable kernel architectures for cross-domain semantic analysis

  • Theoretical analysis of semantic space transferability

  • Efficient implementation for large-scale text processing

Current Progress:

  • ✅ MCMC algorithm design and initial implementation

  • ✅ Theoretical framework for semantic transferability

  • 🔄 Large-scale empirical evaluation across domains

  • ⏳ Optimization for high-performance computing deployment


“Radio Wave Propagation Modeling with SuperDARN: ML-Enhanced Atmospheric Analysis”

Authors: Stephen Mander, et al.
Target Journal: Journal of Atmospheric and Solar-Terrestrial Physics
Status: 50% complete, expected submission June 2025

Abstract: This work integrates machine learning techniques with SuperDARN atmospheric data to develop enhanced radio wave propagation models, enabling improved understanding of ionospheric dynamics and space weather prediction.

Key Innovations:

  • ML-enhanced models for radio wave propagation prediction

  • Integration with global SuperDARN radar network data

  • Real-time atmospheric analysis capabilities

  • Open-source framework for atmospheric data processing

Current Progress:

  • ✅ SuperDARN data integration and preprocessing pipeline

  • 🔄 ML model development for propagation prediction

  • ⏳ Real-time analysis framework implementation

  • ⏳ Validation with historical atmospheric events


Planned Research Extensions

Multi-Year Research Program (2025-2027)

“Scaling Laws in Cross-Modal Systems: A Unified Theory”

Target: Nature Machine Intelligence or Science Advances
Timeline: 18-month development project

Research Vision: Develop comprehensive theoretical framework connecting scaling laws in contrastive learning with cross-modal understanding, bridging my work on 6-dimensional CLIP training with fundamental scaling analysis.

Planned Contributions:

  • Unified mathematical framework for cross-modal scaling analysis

  • Empirical validation across vision-language model families

  • Resource-efficiency guidelines for large-scale multi-modal training

  • Open datasets and evaluation protocols for community use


“Human-AI Optimization: From Tactile Interfaces to Collaborative Intelligence”

Target: ACM Transactions on Interactive Intelligent Systems
Timeline: 24-month collaborative project

Research Vision: Expand tactile optimization work into comprehensive framework for human-AI collaboration in complex problem solving, integrating insights from haptic interfaces with advanced AI reasoning.

Planned Contributions:

  • Theoretical framework for human-AI collaborative optimization

  • Multi-modal interfaces combining tactile, visual, and AI reasoning

  • Evaluation protocols for human-AI team performance

  • Applications across multiple optimization domains


Future Research Directions

Emerging Research Themes

Cross-Modal Foundation Models

Building on current contrastive learning research to develop next-generation foundation models that understand relationships across vision, language, audio, and tactile modalities.

Efficient AI for Edge Computing

Extending 8-bit LSA and tiny contrastive learning work toward comprehensive edge AI frameworks that maintain performance while minimizing computational requirements.

AI Security and Robustness

Advancing PICTAR and adversarial attack research toward comprehensive security frameworks for multi-modal AI systems in critical applications.

Scientific AI Applications

Expanding SuperDARN and atmospheric modeling work toward general frameworks for AI-enhanced scientific discovery and analysis.

Collaboration Opportunities

Seeking Collaborations In:

  • Industry Partnerships: Edge AI deployment and optimization

  • Academic Collaborations: Cross-modal learning theory and applications

  • Interdisciplinary Projects: Human-computer interaction and tactile computing

  • Open Science Initiatives: Reproducible research and open-source frameworks

Publication Strategy & Targets

High-Impact Venues (2025-2026)

  • Nature Machine Intelligence: Scaling laws and theoretical frameworks

  • Science: Cross-modal foundation model breakthroughs

  • ICLR: Advanced contrastive learning methodologies

  • NeurIPS: Optimization and human-AI collaboration

  • CVPR: Vision-language model improvements

Specialized Journals

  • Computational Linguistics: Semantic analysis and NLP innovations

  • AI Magazine: Game AI and strategic reasoning applications

  • Journal of Atmospheric Sciences: Scientific AI applications

Conference Targets

  • ICLR 2025: 6-dimensional contrastive learning framework

  • NeurIPS 2025: Scaling laws in multi-modal systems

  • CHI 2025: Tactile optimization interfaces

  • CVPR 2025: Advanced adversarial robustness techniques

Open Science Commitments

Reproducibility & Accessibility

  • Open Source: All research software publicly available

  • Open Data: Datasets and benchmarks for community use

  • Documentation: Comprehensive tutorials and implementation guides

  • Education: Workshop and tutorial development for research community

Community Engagement

  • Code Reviews: Active participation in open-source ML projects

  • Mentorship: Graduate student and junior researcher guidance

  • Peer Review: Regular reviewer for top-tier conferences and journals

  • Outreach: Public talks and educational content creation


For completed work, see Journal Articles, Conference Papers, and Preprints.

Research software implementations available at Research Software.

Last updated: Sep 16, 2025