NeuroGenesysThe research we build on.
1,514 papers by other researchers that our work cites, from continual learning and verification to language models, agents and hardware. Each entry links to the paper on arXiv and to where we use it. 301 are cited on published pages of this site; the rest in the research notes behind our studies. Our own papers are in the archive.
- 2026-06arXiv 2609.04217At Equal Inference Cost, Multi-Agent Structure Does Not Beat a Single Frozen Agent
David Dylan, Aoife Brennan, Cian Murphy et al. · AI & agents
Multi-agent LLM pipelines, such as Planner-Executor-Critic teams, often report gains over single agents, but these gains usually come with higher inference cost because the team makes multiple model calls per environment step.
- 2026-06arXiv 2606.24026Can Language Model Agents be Helpful Circuit Explainers in Mechanistic Interpretability?
Ayan Antik Khan, Harsh Kohli, Yuekun Yao et al. · AI & agents
Mechanistic interpretability has made substantial progress in automatically localizing circuits, but explaining what localized components do remains labor-intensive and difficult to standardize.
- 2026-06arXiv 2606.24281CALIBER: Calibrating Confidence Before and After Reasoning in Language Models
Conor Finlay, Joshua Kurien, Saurabh Dash et al. · Language models & NLP
Reasoning language models are increasingly asked not only to answer difficult questions, but also to estimate their likelihood of success. Existing methods typically elicit confidence only once: either before thinking or after answering.
- 2026-06arXiv 2606.24322Securing LLM-Agent Long-Term Memory Against Poisoning: Non-Malleable, Origin-Bound Authority with Machine-Checked Guarantees
Yedidel Louck · Security
LLM agents increasingly rely on persistent long-term memory, which creates a critical vulnerability that we study here: memory poisoning.
- 2026-06arXiv 2606.24752Can Scale Save Us From Plasticity Loss in Large Language Models?
J. Fernando Hernandez-Garcia, Tomás Figliolia, Beren Millidge · AI & agents
The loss of plasticity - the ability of a network to learn new information after having already learned older information - is a fundamental challenge in creating artificial neural networks capable of continual learning.
- 2026-06arXiv 2606.24993The Geometry of Sequential Learning: Lie-Bracket Prediction of Transfer Order
John Sweeney · Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:117806-117825, 2026 · Machine learning
possible curricula. We show that the local order effect is governed by a computable geometric quantity, the Lie-bracket commutator of gradient update fields, yielding a pairwise score for whether A->B or B->A is better for a target domain.
- 2026-06arXiv 2607.16227LLM Unlearning for Cyber Defense: A Survey on Methods, Challenges, and Emerging Threats
Ruppikha Sree Shankar, Abhishek Bhardwaj, Arnav Doshi et al. · Machine learning
LLMs are increasingly deployed in security-critical systems across healthcare, finance, education, and decision support, yet their inability to forget creates serious cybersecurity, privacy, and safety risks.
- 2026-06arXiv 2606.23063Attention-Spectrum Regularization for Replay-Free Continual Multimodal LLMs
Chuangxin Zhao, Canran Xiao, Siyuan Ma et al. · Vision
Multimodal large language models (MLLMs) are increasingly required to adapt to non-stationary streams of visual domains, question types, and user instructions, yet continual fine-tuning often causes severe forgetting of previously acquired multimodal skills.
- 2026-06arXiv 2606.23276Exposing the Illusion of Erasure in Knowledge Editing for LLMs
Advik Raj Basani, Anshuman Chhabra · Machine learning
Knowledge Editing (KE) has emerged as a frontier for updating specific facts in LLMs without costly retraining, but its reliability and underlying mechanisms remain poorly understood.
- 2026-06arXiv 2606.22030When Does Belief-Based Agent Memory Help? Reliability-Conditional Updating and Provenance-Capped Poisoning Defense
Pranav Singh · AI & agents
We investigate when belief-based memory actually improves large language model (LLM) agents.
- 2026-06arXiv 2606.19679LOKI: Memory-Free Null-Space Constrained Lifelong Knowledge Editing
Masih Eskandar, Miquel Sirera Perelló, Stratis Ioannidis et al. · Machine learning
Lifelong knowledge editing aims to efficiently and sequentially update language models over time, as new knowledge becomes available or when the model makes mistakes, while preserving acceptable performance on past knowledge.
- 2026-06arXiv 2606.17519Scaling Enterprise Agent Routing: Degradation, Diagnosis, and Recovery
Kellen Gillespie, Robyn Perry · Language models & NLP
Production LLM assistants route user requests to growing libraries of specialized tools, but how does routing accuracy degrade as the catalog scales?
Cited inInternal notes - 2026-06arXiv 2606.17930How Inference Compute Shapes Frontier LLM Evaluation
Jessica McFadyen, Ole Jorgensen, Harry Coppock et al. · AI & agents
AI evaluations are shifting toward harder tasks that benefit from longer trajectories involving tool use and iterative problem solving. As a result, performance is increasingly sensitive to the amount and allocation of compute available at test time ("inference compute").
- 2026-06arXiv 2606.18024Catastrophic Forgetting is Low-Rank: A Function-Space Theory for Continual Adaptation
Ido Nitzan Hidekel, Dan Raviv · Machine learning
Catastrophic forgetting in continual adaptation is usually studied through parameter drift, replay, or distillation, but these views do not identify which output-space directions are vulnerable.
- 2026-06arXiv 2606.17168RepSelect: Robust LLM Unlearning via Representation Selectivity
Filip Sondej, Yushi Yang, Adam Mahdi · Language models & NLP
When LLM weights are open or fine-tuning is available through an API, suppressing hazardous knowledge and tendencies is not enough: removal has to be deep enough that an adversary cannot restore it.
- 2026-06arXiv 2606.17104Prefill/Decode-Aware Evaluation of LLM Inference on Emerging AI Accelerators
Shun Usami, Venkatram Vishwanath, E. Wes Bethel · Hardware, photonics & physics
As large language models (LLMs) are increasingly deployed in latency- and cost-sensitive settings, inference efficiency has become a central systems challenge.
Cited inThe Wrong Phase - 2026-06arXiv 2606.14929Policy Regret for Embedding Model Routing: Contextual Bandits with Low-Rank Experts
Yan Dai, Negin Golrezaei, Patrick Jaillet · Machine learning
Modern recommendation systems increasingly rely on dynamically routing diverse queries to multiple embedding models. Despite its practical significance, this problem remains poorly understood under realistic conditions like adversarial queries, bandit feedback, and limited observability of models.
- 2026-06arXiv 2606.12883The Hidden Power of Scaling Factor in LoRA Optimization
Zicheng Zhang, Haoran Li, Jiaxing Wang et al. · AI & agents
In Low-Rank Adaptation (LoRA), the scaling factor is often treated as a mere complement to the learning rate, yet its role in optimization remains poorly understood.
- 2026-06arXiv 2606.12990Exposure Bias as Epistemic Underidentification in Recursive Forecasting
Riku Green, Zahraa S. Abdallah, Telmo M Silva Filho · Machine learning
Recursive multi-step forecasting is usually framed as distribution shift: models are trained on observed histories but deployed on their own predictions.
- 2026-06arXiv 2606.10554Benchmarking Knowledge Editing using Logical Rules
Tatiana Moteu Ngoli, NDah Jean Kouagou, Hamada M. Zahera et al. · The Semantic Web. ISWC 2025. ISWC 2025. Lecture Notes in Computer Science, vol 16141. Springer, Cham · Language models & NLP
Large Language Models (LLMs) are increasingly deployed in real-world applications that require access to up-to-date knowledge. However, retraining LLMs is computationally expensive.
- 2026-06arXiv 2606.11435Agent Skill Evaluation and Evolution: Frameworks and Benchmarks
Kexin Ding, Yang Zhou, Can Jin et al. · Language models & NLP
The growth of agent skills has transformed how agentic systems are built, evaluated, and deployed. As skill libraries continue to scale, rigorous evaluation becomes critical to ensuring their utility, quality, and safety in real-world applications.
Cited inStudy: Genesys program prior art - 2026-06arXiv 2606.09707BrainSurgery: Reproducible and Flexible Declarative Weight Manipulations for Model Editing and Upcycling
Gianluca Barmina, Annemette Broch Pirchert, Andrea Blasi Núñez et al. · Machine learning
Checkpoint transformations such as model merging, precision casting, and low-rank rewriting are commonly implemented through Python code or separate task-specific tools.
- 2026-06arXiv 2606.09932When RL Fails after SFT: Rejuvenating Model Plasticity for Robust SFT-to-RL Handoff
Runze Liu, Jiashun Liu, Xu Wan et al. · Machine learning
Supervised Fine-Tuning (SFT) followed by Reinforcement Learning (RL) has become a standard pipeline for Large Language Model (LLM) post-training. SFT is expected to provide a useful behavioral prior for RL to further enhance model capabilities.
- 2026-06arXiv 2606.09899When Attribution Patching Lies: Diagnosis and a Second-Order Correction
Luyang Zhang, Jialu Wang · Machine learning
A central goal of mechanistic interpretability is to identify which internal components causally drive a language model's behavior. Because these importance estimates serve as the evidence for identifying circuits, systematic errors can lead to the misidentification of the underlying mechanisms.
- 2026-06arXiv 2606.05661Continual Learning Bench: Evaluating Frontier AI Systems in Real-World Stateful Environments
Parth Asawa, Christopher M. Glaze, Gabriel Orlanski et al. · AI & agents
Continual learning, the ability of AI systems to improve through sequential experience, has attracted substantial interest, but no high-quality benchmark exists to evaluate it.
- 2026-06arXiv 2606.04923Reproducing, Analyzing, and Detecting Reward Hacking in Rubric-Based Reinforcement Learning
Xuekang Wang, Zhuoyuan Hao, Shuo Hou et al. · Machine learning
Rubric-based reinforcement learning (RL) uses an LLM-as-a-Judge (LaaJ) to score model outputs according to rubrics as rewards. However, policy models may exploit latent biases in the judge, leading to reward hacking and ineffective or unsafe training outcomes.
Cited inInternal notes - 2026-06arXiv 2606.03179HyperPatch: Sequential Knowledge Editing Under n-ary Structural Drift
Yu-Kai Chan, Wen-Sheng Lien, Dong-Ting Yao et al. · Language models & NLP
Large Language Models (LLMs) rely on Knowledge Editing (KE) to maintain temporal validity, yet real-world knowledge is inherently n-ary.
- 2026-06arXiv 2606.03695Don't Forget Your Embeddings: Robust Knowledge Erasure via Precise Editing of Embeddings
Clara Haya Suslik, Or Shafran, Mor Geva · Language models & NLP
As language models are increasingly deployed in real-world applications, the ability to erase specific knowledge from them becomes critical for safety and compliance.
- 2026-06arXiv 2606.03979Language Models Need Sleep: Learning to Self-Modify and Consolidate Memories
Ali Behrouz, Farnoosh Hashemi, Adel Javanmard et al. · Machine learning
The past few decades have witnessed significant advances in the design of machine learning algorithms, from early studies on task-specific shallow models to more general deep Large Language Models (LLMs).
Cited inInvestors are paying for agents that learn on the job. The hard part is knowing what they forgot.To keep what a model knows, control what it is allowed to moveWhat a model lets into long-term memory matters more than how it learnsStudy: Fortifying knowledge in continual learningStudy: Genesys program prior artStudy: Hardware guardian core for continual learningStudy: LLMs modifying open model weightsStudy: Logical coherence of LLM world modelsStudy: Memory world models and source trustStudy: Seven continual learning methods tested - 2026-06arXiv 2608.12332Can Spectral-Clipping Enable Better Learning While Forgetting Less for Low-Rank Adaptation?
Hyowon Wi, Noseong Park · Language models & NLP
In recent years, low-rank adaptation (LoRA) has emerged as a significant paradigm that freezes pre-trained weights and introduces small, learnable adapters instead of fine-tuning the full set of parameters.
- 2026-06arXiv 2606.01610Revisiting Ripple Effects in Knowledge Editing through Pressure-Aware Joint Neighborhood Optimization
Haoben Huang, Shuxin Liu, Ou Wu et al. · AI & agents
Single-edit updates in large language models can trigger ripple effects across local knowledge neighborhoods: desirable propagation to related facts and unintended perturbation of preserved ones.
- 2026-06arXiv 2606.02800Cosmos 3: Omnimodal World Models for Physical AI
NVIDIA, :, Aditi et al. · Vision
We introduce Cosmos 3, a family of omnimodal world models designed to jointly process and generate language, image, video, audio, and action sequences within a unified mixture-of-transformers architecture.
- 2026-06arXiv 2606.09863From Confident Closing to Silent Failure: Characterizing False Success in LLM Agents
Laksh Advani · Machine learning
LLM agents can fail silently by asserting task completion when the environment state shows otherwise.
Cited inInternal notes - 2026-05arXiv 2606.00820Not All Flips Are Conformity: Decomposing Stance Convergence in Multi-Agent LLM Debate
Xiqi Hao, Zengqing Wu, Yu-Xuan Qiu et al. · Language models & NLP
Multi-agent debate (MAD) among large language models (LLMs) is often judged by answer changes and convergence, yet these outcomes alone cannot establish that peers improved the answer.
- 2026-05arXiv 2605.31025TRACE: Discovering Task-Specific Parameter via Adaptation-Aware Probing for Continual Fine-Tuning
Xiaosong Han, Ke Chen, Xindi Dai et al. · Language models & NLP
In real-world deployment, LLMs are often adapted continually across tasks to keep LLMs up-to-date in production, where new fine-tuning should preserve previously learned skills.
- 2026-05arXiv 2605.31268Mellum2 Technical Report
Marko Kojic, Ivan Bondyrev, Aral de Moor et al. · Language models & NLP
5B active parameters per token.
- 2026-05arXiv 2605.31314AR Forcing: Towards Long-Horizon Robot Navigation World Model
Yifei Yang, Zehua Fan, Huan Li et al. · Robotics & control
The diffusion based robot navigation world models are typically trained using parallel supervision, while autoregressive inference is employed during path planning.
- 2026-05arXiv 2606.24893AgentOdyssey: Open-Ended Long-Horizon Text Game Generation for Test-Time Continual Learning Agents
Zheyuan Zhang, Zehao Wen, Alvin Zhang et al. · Language models & NLP
For agents to learn continuously from interaction with the world at test time, they must be able to explore effectively, acquire new world knowledge and skills, retain relevant episodic experiences, and plan over long horizons.
Cited inInternal notes - 2026-05arXiv 2608.14567From Doyle to AGM: A Survey and an Implementation Roadmap for Belief Change
Yuri Almeida, Arthur Casals · AI & agents
This paper presents a targeted narrative review establishing the historical and theoretical foundations for computational belief change implementation.
- 2026-05arXiv 2605.29350ConMoE: MoE Compression as Expert Selection and Functional Assignment
Yilun Yao, Jiaming Pan, Elsie Dai et al. · AI & agents
Sparse Mixture-of-Experts (MoE) models scale model capacity through sparse activation, but storing their full parameters remains costly in GPU memory.
Built from every arXiv paper cited in our papers, posts, topic hubs, reports and research notes, with titles, authors, dates and abstracts from the arXiv API (updated 2026-11-12). Summaries are the opening sentences of each paper's own abstract, not our assessment; a citation means we used or checked the work, not that we endorse it. Fields follow each paper's primary arXiv category. Papers cited only outside arXiv (journals, proceedings) are not listed yet. Entries are generated from our citations, so a mistyped ID in a note would list the wrong paper; we remove those as we find them. Report a wrong entry.
