Latest Complete Week
863
2026-07-13
Cognaptus DataHub Monitor
arXiv AI Research Monitor
A monitored view of publication volume, topic momentum, weekly keywords, and selected papers across arXiv AI research categories.
Weekly Change
-42
papers versus prior complete week
Tracked Topics
12
subject-topic series available
Featured Papers
11
in the latest reporting period
What This Monitor Shows
This monitor is a structured view of selected arXiv categories, not a measure of all AI research. It is designed to make changes in research volume and topical attention easier to inspect over time.
The data is refreshed by an external research-processing workflow. Each panel states its coverage and remains available when other panels have not yet been generated.
Long-Run Signal
Monthly cs.AI Publication Volume
Published papers by complete calendar month. The visual keeps the full series available from the configured start date.
Monthly cs.AI volume uses only ref_arxiv_cs_ai.fst. Submission dates are inferred from bare arXiv identifiers; days beyond month end are capped. The current incomplete month is excluded.
Monthly Signal
Topic Momentum by Subject
Each series is supplied by the external topic classifier. Use the subject and topic labels as directional research signals, not fixed disciplinary boundaries.
cs.AI
Foundation models and pretraining
2016-01 to 2026-06
cs.AI
Post-training and alignment
2016-01 to 2026-06
cs.AI
Reasoning, planning, and agentic AI
2016-01 to 2026-06
cs.AI
Retrieval, memory, and knowledge grounding
2016-01 to 2026-06
cs.AI
Multimodal vision, audio, video, and documents
2016-01 to 2026-06
cs.AI
Embodied AI, robotics, and world models
2016-01 to 2026-06
cs.AI
Efficient training, inference, and AI systems
2016-01 to 2026-06
cs.AI
Evaluation, benchmarks, and reliability
2016-01 to 2026-06
cs.AI
Safety, security, privacy, and governance
2016-01 to 2026-06
cs.AI
Data-centric AI and synthetic data
2016-01 to 2026-06
cs.AI
AI for science, medicine, and engineering
2016-01 to 2026-06
cs.AI
Business, social applications, and human-AI workflows
2016-01 to 2026-06
Current Attention
Top cs.AI Keywords This Week
2026-07-13 to 2026-07-19
Counts refer to papers containing each normalized keyword or phrase.
VLMs
19
2.2%
MLLMs
19
2.2%
Explainability
14
1.6%
mixture of experts
14
1.6%
Transformers
15
1.7%
GRPO
12
1.4%
Mathematical Reasoning
11
1.3%
Hallucination
15
1.7%
LoRA
15
1.7%
Code Generation
11
1.3%
Human-Readable Starting Points
Featured Papers
Selected records supplied by the research workflow for the latest reporting period.
2026-07-14
Deconstructing Actor-Critic: A Large-scale Empirical Study of Design Components for Practitioners
Top lexical match for AI for science, medicine, and engineering in the latest complete cs.AI week.
cs.LGcs.AI
2026-07-15
Adversarial Prompting Framework for AI Safety Assessment
Top lexical match for Business, social applications, and human-AI workflows in the latest complete cs.AI week.
cs.CRcs.AI
2026-07-14
Privacy Preserving Recommender Systems Balancing Personalization with Privacy
Top lexical match for Data-centric AI and synthetic data; Safety, security, privacy, and governance in the latest complete cs.AI week.
cs.CRcs.AIcs.LG
2026-07-14
Efficient and Privacy Aware Edge Cloud Collaborative Inference for Large Language Models
Top lexical match for Efficient training, inference, and AI systems in the latest complete cs.AI week.
cs.CRcs.AI
2026-07-14
TerraZero: Procedural Driving Simulation for Zero-Demonstration Self-Play at Scale
Top lexical match for Embodied AI, robotics, and world models in the latest complete cs.AI week.
cs.LGcs.AIcs.RO
2026-07-15
Set-shifting Behavioral Test for Harnessed Agents
Top lexical match for Evaluation, benchmarks, and reliability in the latest complete cs.AI week.
cs.AIcs.CLcs.SE
2026-07-15
MxGPS: Multiplex Graph Transformers for a Power Grid Foundation Model
Top lexical match for Foundation models and pretraining in the latest complete cs.AI week.
cs.LGcs.AI
2026-07-15
GHR-VLM: Making Zero-Shot Transit Video Analytics Realizable with Grounded Hybrid Reasoning
Top lexical match for Multimodal vision, audio, video, and documents in the latest complete cs.AI week.
cs.CVcs.AI
2026-07-15
Groc-PO: Grounded Context Preference Optimization for Truthful Multimodal LLMs
Top lexical match for Post-training and alignment in the latest complete cs.AI week.
cs.CVcs.AIcs.CLcs.MM
2026-07-15
The Dynamic Verifiable Multi-Agent Human Agentic Loyalty Loop (DVM-HALL) Model and the Net Human-Agent Score (NHAS) in Autonomous Commerce
Top lexical match for Reasoning, planning, and agentic AI in the latest complete cs.AI week.
cs.SIcs.AIcs.GTcs.MA
2026-07-14
RAGthoven at SemEval-2026 Task 1: A Multi-Stage Pipeline Walks Into a Benchmark and Barely Clears the Bar
Top lexical match for Retrieval, memory, and knowledge grounding in the latest complete cs.AI week.
cs.CLcs.AI
Methodology and coverage
Source: LitXR local arXiv category corpora. Open source
Categories: cs.AI, cs.CL.
Monthly publication volume uses only the cs.AI thin index. Identifier-derived dates are aggregated by month, and the current incomplete month is excluded.
Topic momentum uses LitXR's ai_category_keywords_v1 lexical query set. A paper can match multiple categories, and only papers meeting the configured distinct-phrase threshold are counted.
A subject appears in topic momentum only after LitXR has built that collection's abstract lexical metadata and postings index.
Weekly keywords are normalized tags established in prior Cognaptus articles. Explicit stoplist terms and tags appearing in more than 10% of articles are excluded. Remaining terms rank by natural-log weekly lift multiplied by weekly paper count.
Featured papers are the highest-scoring current-week cs.AI paper in each lexical category. Canonical JSON is hydrated only after this bounded selection to obtain display metadata.