PhD Doctoral Papers Pipeline & Integrity Dashboard

Automated Cross-Paper Audit, Hardware Smaps Rollup Telemetry & AGENTS.md Rule 5 Guardian

Pipeline v1.2.0 • Sync: Active
🛡️
AGENTS.md Rule 5 Compliance: PASSED (100% In-Step) Zero Overlap Verified
REPORTED_CAMPAIGNS (7) and EXCLUDED_CAMPAIGNS (4) are strictly synchronized across python scripts and numbers.tex. Table-to-text sum desynchronization prevented.

Live Edge Telemetry: Experiment D (Extended 48h Mechanism Isolation & Multi-Horizon Rejuvenation)

ubuntu-llm (Intel Core i5-8250U CPU-only Edge Node, Kaby Lake-R)
Inferences Completed 21,840 Burn-in: 16,137 inf (24h)
VmRSS Memory 3703.1 MB Initial: 3658.3 MB (+44.8 MB)
Private Dirty Heap 98.5% Dirty: 3675.5 MB (smaps_rollup)
Decode Speed 24.46 tps Aging Decay: -1.22%
Active Phase & 3-Layer Isolation Results:
• Current State: Phase 3: Post-Intervention (L2 Page Cache Drop - POSIX_FADV_DONTNEED / drop_caches=3)
• L1 (Thermal Relaxation): Idle 60s paused inference → 0.5% RSS recovery (-0.24 MB). Thermal decay rejected.
• L2 (OS Page Cache Drop): Cleared OS page caches via POSIX_FADV_DONTNEED → 0% recovery. Proves memory accumulation is 100% heap-bound.
• L3 (Runner Restart Rejuvenation): Instant RSS reset with energy cost ΔErejuv = 84.7 J, directly calibrating Paper 4's closed-form model.

Active Doctoral Papers Pipeline

PAPER1-IOTJ IEEE Internet of Things Journal (IoT-J)
Submitted & Under Review

Slowing Down to Save: A Longitudinal, Carbon-Aware Evaluation of Large Language Model Inference on CPU-Only Edge Nodes

Covers 11 longitudinal 48h campaigns; establishes Pareto-optimal CPU frequency scaling; prevents thermal runaway on edge nodes.

Venue Limit 12 pages
Current Size 12 pages
TBD Markers 0
Overfull Hbox 0
PAPER2-AGING-POSITION IEEE Trans. on Emerging Topics in Computing (TETC) SI WoSAR 2026 (Q1, IF 4.8) | AISAR Synergy
Ready for Review (Camera-Ready Alignment)

Software Aging and Rejuvenation for Large Language Model Inference on IoT and Edge Nodes: A Position and Research Agenda

First systematic position defining Aging-Related Bugs (ARBs) in edge-native LLM runtimes (llama.cpp, Ollama, ONNX Runtime).

Venue Limit 6 pages
Current Size 5 pages
TBD Markers 0
Overfull Hbox 0
PAPER3-MECHANISM IEEE/IFIP DSN 2027 Disrupt Track (Berlin) — dsn2027-berlin.github.io
Draft Complete (Exact 7 Pages, Zero Overfull)

Layer Localization and Mechanism Isolation of Performance Decay in Sustained Edge LLM Inference

Dissected 3 layers (hardware thermal, OS page cache, runtime heap). Confirmed 98.5% Private_Dirty heap localization via smaps_rollup; isolated ARB-MEM-01 & ARB-CORO-01.

Venue Limit 7 pages
Current Size 7 pages
TBD Markers 0
Overfull Hbox 0
PAPER4-FRAMEWORK IEEE Trans. Green Communications and Networking (TGCN) / IEEE TSE
Architecture Formulated & Calibrated

Energy- and Carbon-Aware Software Rejuvenation Scheduling for Edge LLM Services

EcoSAR optimal rejuvenation scheduling policy: closed-form T* balancing degradation carbon penalty against rejuvenation energy overhead Delta E_rejuv.

Venue Limit 12 pages
Current Size Formulation Stage
TBD Markers 0
Overfull Hbox 0

Exposed Pipeline Tools & Direct API

This Cloudflare worker provides both an interactive view and a Model Context Protocol endpoint for autonomous agents:

GET /api/pipeline Full JSON pipeline state & telemetry snapshot POST /mcp Model Context Protocol JSON-RPC 2.0 tool interface