# v0.9.10 → v0.9.11 — Research Expert Team Deep Expansion (Oct 15, 2026, part 2)

## Summary

**Pure skill-content release — no carrier changes**: 955 methods / 26 categories / 18 skills all unchanged. `research-expert-team` grew from a nine-role router into a **two-layer methodology asset** (program + evidence) with a 3DGS paper-drafting blueprint, expert-review audited and refreshed against 2025-2026 literature. Detailed work log: [changelog/2026-10-15.md](2026-10-15.md) (Waves 1-5).

## Skill Expansion: research-expert-team (references/, 1 → 7 files)

- **`references/methodology-core.md` (NEW, program layer)** — executable procedures distilled from 15 top-cited exemplars, domain-tuned to 3DGS/spatial-intelligence:
  - §A topic-selection engine: modes M1-M4 + five opening-sentence variants A.1 (clash / paradox / agenda-revival / pipeline-dissection / legacy-gap), red-flag checklist, delta typology (5 power levels)
  - §B argumentation architecture: skeleton templates T1-T3, contribution-evidence bijection, §B.5 representation-paper techniques (math anchor, bounded first-claim, motivation-upgrade rewriting, symbol-system unification, counter-intuitive preempt), §B.6 domain-skeleton router
  - §C experiment pattern library C1-C16 (dual operating points, mechanism-tiered ablations, Pareto sweeps, single-model zero-finetune multi-task protocol, evaluation-hygiene clauses, cross-method component evidence)
  - §D reviewer-expectation library D1-D8 with preempt actions; §E nine value amplifiers; §F team conventions (mode-number tagging for traceability)
- **`references/mp-*.md` (NEW ×5, evidence layer)** — full-text dissections of 15 exemplars across 5 directions (citations verified via Semantic Scholar batch API + OpenAlex cross-check):
  - mp-ontology.md: Kimera (IJRR 2021, 407) · ConceptGraphs (ICRA 2024, 671)
  - mp-super-agents.md: ReAct (ICLR 2023, 12,540) · MetaGPT (ICLR 2024 Oral, 2,626) · Toolformer (NeurIPS 2023, 6,146) · Generative Agents (UIST 2023, 5,987) · Voyager (TMLR 2024, 2,622)
  - mp-aigc.md: DreamFusion (ICLR 2023, 3,929) · DreamGaussian (ICLR 2024, 1,098)
  - mp-3dgs.md: 3DGS (TOG 2023, 10,887) · 4DGS (CVPR 2024, 1,676) · SuGaR (CVPR 2024, 982)
  - mp-spatial.md: NeRF (ECCV 2020, 5,900) · Instant-NGP (TOG 2022, 6,456) · DUSt3R (CVPR 2024, 2,022)
- **`references/3dgs-paper-blueprint.md` (NEW, domain extension)** — 3DGS-specific drafting skeleton: three writing rules (motivation upgrade to avoid "tuning-paper" feel / absolute symbol unification delegating to cg-paper-writing's symbol table / painpoint→resource-metric mapping with a 3DGS inherent-defect library), five-section placement table, limitations library, cross-asset mapping table, pre-submission checklist — fused with `cg-paper-writing` (no duplicated symbol tables or section templates)
- **`SKILL.md`** — §2.4 rewritten as the two-layer load-by-need/load-by-direction router; role playbooks bind R1→§A, R3→§D, R6→§B(+blueprint), R8/R9→§C; frontmatter description updated with 3DGS-drafting trigger phrases

## Quality Passes (same day)

- **Wave 4 expert-review audit** — 10 papers re-fetched from ar5iv, ~60 numeric claims line-verified: 5 factual errors fixed (INGP-vs-NGLOD IoU cross-scene splice; unverifiable NeRF award claim removed; 3DGS §7.3 subsection count; six-column Table 1; 4DGS Neu3D caliber pivot), 5 imprecisions corrected, SKILL.md "three directions" vs five-direction inconsistency resolved; award claims re-verified (INGP SIGGRAPH 2022 best paper ✓, VGGT CVPR 2025 best paper ✓)
- **Wave 5 latest-paper review** — five-direction 2025-2026 progress searches: stale claims updated (NeRF word-root count, VGGT "hundreds of views" caliber, DGGT lineage), 2026-10 landscape notes added per direction (3DGS deployment bottlenecks / Deep Research Agents / TRELLIS-Hunyuan3D-2 large-model route / TPAMI-level scene graphs); citation-date and venue calibers unified

## Carrier Integration (README + version calibers)

- `README.md` / `README_CN.md`: v0.9.11 What's New / 最新动态 entry (v0.9.10 demoted to Previous); skills-table row 18 description extended (two-layer methodology core + blueprint)
- **Badge caliber bug fixed**: skills badge was stale at 17 (EN) / 15 (CN) — both corrected to **18** (pre-existing miss from v0.9.9/v0.9.10)
- Version 0.9.10 → 0.9.11: `package.json`, `CLAUDE.md`, `.claude-plugin/marketplace.json`, `docs/app-common.js` (hero.version zh+en), `docs/index.html`
- Carrier counts (955 / 26 / 18) intentionally unchanged — no knowledge-base or skill-count deltas; historical records (changelog/, README update-history) untouched
- TeleAgent skill directory (`~/.config/TeleAgent/.../skills/research-expert-team/`) + `~/.workbuddy/skills/` synced in the same passes — 8 files MD5-identical across all three copies

## Verification

- All 15 exemplar papers' facts drawn from ar5iv full-text reads; citation counts Semantic Scholar batch API 2026-10-15 (NeRF via title-match endpoint + OpenAlex cross-check)
- Wave 4 audit: 10 papers re-fetched, ~60 numeric claims line-verified, zero unresolved discrepancies after fixes; stale-claim residual scan clean
- Three-copy MD5 verification (project / workbuddy / TeleAgent): 8/8 files identical
- No carrier changes: `data/methods.json` untouched (955); skill count unchanged (18)

## State

Uncommitted in working tree — user commits manually, per established workflow.