Research library
Search papers on equipment connectivity, quality inspection, production analysis, and manufacturing work.
Search and filter
Paper list
T01. LLM agent orchestration and harness design
External research on how multiple language-model agents are planned, routed, and evaluated.
Show this topic only- T01-14Candidate approachSince 2025
Multi-Agent Collaboration via Evolving Orchestration
- T01-4Candidate approach
Generative Agents: Interactive Simulacra of Human Behavior
- T01-5Candidate approach
Reflexion: Language Agents with Verbal Reinforcement Learning
- T01-6Candidate approach
CAMEL: Communicative Agents for "Mind" Exploration of Large Language Model Society
T02. Recursive language models and long-context handling
Research on reading documents that do not fit in one context window.
Show this topic only- T02-4Candidate approach
MemGPT: Towards LLMs as Operating Systems
- T02-8Candidate approach
LongLLMLingua: Accelerating and Enhancing LLMs in Long Context Scenarios via Prompt Compression
- T02-10Candidate approach
Efficient Streaming Language Models with Attention Sinks (StreamingLLM)
- T02-12Candidate approach
Ring Attention with Blockwise Transformers for Near-Infinite Context
- T02-13Candidate approach
Extending Context Window of Large Language Models via Positional Interpolation
- T02-14Candidate approach
Leave No Context Behind: Efficient Infinite Context Transformers with Infini-attention
- T02-15Candidate approach
H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models
T03. Tool use, function calling, and the Model Context Protocol
External research and specifications for letting a model call outside tools.
Show this topic only- T03-8Candidate approach
ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs
T04. Planning, reflection, judging, and self-improving agents
Research on step-by-step reasoning, self-critique, and model-as-judge evaluation.
Show this topic only- T04-8Candidate approach
G-Eval: NLG Evaluation using GPT-4 with Better Human Alignment
- T04-10Candidate approach
Agent-as-a-Judge: Evaluate Agents with Agents
- T04-12Candidate approach
Self-Rewarding Language Models
- T04-13Candidate approach
Voyager: An Open-Ended Embodied Agent with Large Language Models
- T04-14Candidate approach
STaR: Bootstrapping Reasoning With Reasoning
T05. Human review and approval workflows
Human-in-the-loop and approval research, including automation-induced complacency.
Show this topic only- T05-22Candidate approachSince 2025
What You Approve Is What Executes: Consent Integrity for Black-Box LLM Agents
- T05-23Candidate approachSince 2025
Oversight Has a Capacity: Calibrating Agent Guards to a Subjective, Fatiguing Human
- T05-24Candidate approach
Deep reinforcement learning from human preferences
- T05-25Candidate approach
Human-in-the-loop machine learning: a state of the art
T06. RAG, knowledge graphs, ontologies, and provenance
Retrieval-augmented generation and structured knowledge with traceable sources.
Show this topic only- T06-6Candidate approach
HippoRAG: Neurobiologically Inspired Long-Term Memory for Large Language Models
T07. Structured output, schema validation, and document understanding
Research on forcing machine-checkable output and reading business documents.
Show this topic only- T07-4Candidate approach
Nougat: Neural Optical Understanding for Academic Documents
T08. OCR, vision-language models, and industrial display reading
Research on reading meters, indicators, and shop-floor displays from images.
Show this topic only- T08-14Candidate approachSince 2025
Qwen2.5-VL Technical Report
- T08-7Candidate approach
TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models
- T08-11Candidate approach
OCRBench: On the Hidden Mystery of OCR in Large Multimodal Models
- T08-13Candidate approach
Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution
T09. Industrial time series, anomaly detection, and predictive maintenance
Sensor-driven fault detection, remaining useful life, and condition monitoring research.
Show this topic only- T09-16Candidate approach
Deep learning models for predictive maintenance: a survey, comparison, challenges and prospects
T10. AutoML, model selection, evaluation design, and experiment tracking
Research on choosing and validating models instead of shipping a single fitted model.
Show this topic only- T10-4Candidate approach
Hyperband: A Novel Bandit-Based Approach to Hyperparameter Optimization
- T10-5Candidate approach
BOHB: Robust and Efficient Hyperparameter Optimization at Scale
- T10-6Candidate approach
Optuna: A Next-generation Hyperparameter Optimization Framework
- T10-15Candidate approach
Model Cards for Model Reporting
- T10-16Candidate approach
Datasheets for Datasets
T11. Object detection, multi-object tracking, and video understanding
Detection and tracking backbones behind camera-based safety and inspection work.
Show this topic only- T11-6.3Candidate approachSince 2025
SAM 2: Segment Anything in Images and Videos
- T11-2.7Candidate approach
DETRs Beat YOLOs on Real-time Object Detection (RT-DETR)
- T11-2.8Candidate approach
Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection
- T11-2.9Candidate approach
YOLOv10: Real-Time End-to-End Object Detection
- T11-3.4Candidate approach
Observation-Centric SORT: Rethinking SORT for Robust Multi-Object Tracking (OC-SORT)
- T11-5.3Candidate approach
VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training
T12. Video understanding and evidence selection with vision-language models
Research on explaining what a camera saw and pointing back to the evidence frame.
Show this topic only- T12-5Candidate approach
A Simple LLM Framework for Long-Range Video Question-Answering (LLoVi)
T13. Edge AI, streaming inference, and resource scheduling
Running models near the machine under limited compute and latency budgets.
Show this topic only- T13-9Candidate approach
INFaaS
- T13-10Candidate approach
Orca: A Distributed Serving System for Transformer-Based Generative Models
- T13-11Candidate approach
Efficient Memory Management for Large Language Model Serving with PagedAttention
- T13-12Candidate approach
Gandiva: Introspective Cluster Scheduling for Deep Learning
- T13-13Candidate approach
Tiresias: A GPU Cluster Manager for Distributed Deep Learning
- T13-14Candidate approach
AntMan: Dynamic Scaling on GPU Clusters for Deep Learning
- T13-15Candidate approach
Heterogeneity-Aware Cluster Scheduling Policies for Deep Learning Workloads
- T13-16Candidate approach
Pollux: Co-adaptive Cluster Scheduling for Goodput-Optimized Deep Learning
T14. Industrial protocol translation, code generation, and program synthesis
Research on generating and checking the code that talks to plant equipment.
Show this topic only- T14-4Candidate approach
Agents4PLC: Automating Closed-loop PLC Code Generation and Verification in Industrial Control Systems using LLM-based Agents
- T14-8Candidate approach
Discoverer: Automatic Protocol Reverse Engineering from Network Traces
- T14-9Candidate approach
NetPlier: Probabilistic Network Protocol Reverse Engineering from Message Traces
- T14-14Candidate approach
Polyglot: Automatic Extraction of Protocol Message Format using Dynamic Binary Analysis
T15. OPC UA, Asset Administration Shell, MQTT, and manufacturing interoperability
Specifications and research for describing equipment and moving its data.
Show this topic only- T15-2.6Candidate approach
Open-Source Implementations of the Reactive Asset Administration Shell: A Survey
- T15-2.7Candidate approach
Generation of Asset Administration Shell With Large Language Model Agents: Toward Semantic Interoperability in Digital Twins in the Context of Industry 4.0
T16. Manufacturing knowledge graphs and semantic layers
Research on giving plant data a shared meaning across systems.
Show this topic only- T16-8Candidate approachSince 2025
Intent-Driven Smart Manufacturing Integrating Knowledge Graphs and Large Language Models
- T16-4Candidate approach
Literal-Aware Knowledge Graph Embedding for Welding Quality Monitoring: A Bosch Case
T17. Natural language to SQL and grounded report generation
Research on turning a question into a checked query and a sourced report.
Show this topic only- T17-19Candidate approach
InfiAgent-DABench: Evaluating Agents on Data Analysis Tasks
T18. AI quality management, model monitoring, and audit trails
Research and standards for keeping a deployed model accountable over time.
Show this topic only- T18-11Candidate approachSince 2025
Time to Retrain? Detecting Concept Drifts in Machine Learning Systems
From research to product use
Operating capabilities, pilots, and technologies in development are identified separately.
View technology