This static portal bundles the literature-review pipeline, the error-analysis synthesis workflow, and the main interactive outputs for reviewing human-agent collaboration research gaps.
This site is intended as a review aid for discussing possible research questions, not as a final literature-review chapter. The goal is to make the search process, intermediate evidence, and synthesized challenge landscape inspectable in one place.
Suggested route: start with the two pipeline diagrams to understand how papers were collected and how failure evidence was extracted; then open the 3D research space for the paper-level landscape; then use the challenge visual space for temporal/community patterns; finally audit any claim through the evidence browser, which preserves the extracted text, paper metadata, and section/page hints.
The cluster and trend views are synthesis tools. They are useful for spotting recurring problems, emerging gaps, and community-specific framings, but individual claims should be checked against the evidence rows before being treated as research findings.
High-level architecture for the corpus workflow: venue harvesting, lexical prefiltering, LLM triage, manual review, and Stage 5 synthesis.
Process map for extracting failure evidence from papers, merging full-text and candidate-chunk signals, normalizing patterns, and producing challenge-level analysis.
Interactive 3D map of the retained Stage 5 papers with roles, benchmark markers, citation overlays, paper details, and trend-oriented views of the research landscape.
Full merged evidence browser for auditing extracted limitations, failure modes, result breakdowns, raw quotes, sections, pages, paper metadata, and extraction provenance.
Interactive challenge synthesis view with cluster definitions, semantic map, yearly dynamics, Sankey-style event/community flow, and evidence tracebacks.
Deploy the whole advisor_site/ directory as a static website. The bundled pages use relative links,
and the copied Plotly/Three.js assets are stored locally under assets/.
Local preview command: python3 -m http.server 8088 --directory docs/advisor_site