Research Projects
Scope
This topic tracks Vipin’s active research projects and their durable claim boundaries.
Project Clusters
- AI4EDA / analog design agents: analog-agent
- LLM-based recommendation uncertainty: uncertainty, truce-rec
- Temporal recommendation evidence: tgl-rec
- Agent benchmark / specification grounding: donebench
- AI4S protein uncertainty: uncertaintyprotein-ai4s, protein-optimization-feedback-shift
- Paper writing and literature-mapping workflow support: paper-orchestra
Cross-Cutting Themes
- EXTRACTED: Several projects use strict evidence labels and forbid paper claims before audits/provenance pass.
- EXTRACTED: Recommendation projects converge on same-candidate score schemas and official-baseline governance.
- EXTRACTED: AI4EDA and AI4S projects both study closed-loop decision systems where uncertainty must remain useful under distribution shift or expensive truth queries.
- INFERRED: Vipin’s research portfolio is increasingly about decision-relevant uncertainty, not uncertainty as a passive prediction metric.
- INFERRED: For future research ideation, the wiki should support aggressive reframing and research-ideation-policy, not only summarizing the current project state.
- INFERRED: paper-orchestra can support project write-ups by converting verified project material into outlines, literature reviews, figures, LaTeX sections, and refinement loops; it should not replace project-specific evidence checks.
Counterpoints and Gaps
- The current wiki pass is a high-level project digest, not a full ingest of every draft, PDF, result table, or server output.
- Many claims remain status-bound to each project’s own docs; final paper claims require deeper per-project evidence checks.
- Public wiki pages should avoid copying sensitive raw outputs, credentials, private server paths, or large artifact details.