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.
- 2026-11arXiv 2611.14326When Should Agents Remember? Falsification-Gated Self-Evolution for LLM Agents
Runxuan Tang, Haoyu Gao, Yuyan Ding et al. · AI & agents
LLM agents increasingly operate in long-horizon, verifier-rich environments, but current self-improvement mechanisms often write episode-derived lessons directly into memory, making local success an unreliable signal for reusable and safe knowledge.
- 2026-11arXiv 2611.12240Prune, Don't Merge: Expert Compression of Mixture-of-Experts LLMs on Multilingual Production Tasks
Tomas Rasymas, Povilas Kvedaras · Language models & NLP
Mixture-of-Experts language models are cheap to run per token but expensive to hold in GPU memory, which drives the cost of serving them in production. One-shot expert compression shrinks them after a single calibration pass.
- 2026-11arXiv 2611.12953Agent Architecture Cost-Success Trade-offs: A Controlled Evaluation across Enterprise-Style SOPs
Subhrangshu Nandi, Hitesh Arora, Yunfei Bai et al. · AI & agents
Enterprises increasingly use large language model agents to execute standard operating procedures (SOPs): long, branching, tool-intensive workflows. Yet there is little controlled evidence for choosing among competing agentic architectures.
- 2026-11arXiv 2611.05919Routes of Reasoning: Reading and Reshaping Test-Time Compute in MoE Models
Shengjie Wang, Mohsen Fayyaz, Nanyun Peng · AI & agents
Reasoning models allocate test-time compute across distinct behaviors such as planning, deriving, and verifying. However, models frequently "overthink," continuing to verify and re-derive long after finding a correct answer.
- 2026-11arXiv 2611.06760Route First, See Later: Enabling Vision Agents to Reason Scientifically by Separating Routing Policy from Action Spaces
Shubhabrata Mukherjee, Jeffrey Donatelli, Daniela Ushizima · Vision
Scientific image-analysis workflows require agents to select among diverse capabilities like description, detection, counting, and segmentation.
- 2026-11arXiv 2611.06858CompressionBench: How can your coding agents compress LLMs?
Xinrui Zhong, Zhifeng Li, Zhiming Wang et al. · AI & agents
Recursive self-improvement requires AI systems to improve the models and infrastructure used by subsequent systems. LLM compression offers a concrete efficiency objective: retain a pretrained model's behavior within a smaller model storage budget.
- 2026-11arXiv 2611.07140More Agents, More Reliable? System-Level Reliability of Multi-Agent LLMs under Operational-Profile Uncertainty
Burkay Tunçtürk, Koorosh Aslansefat, Yiannis Papadopoulos et al. · AI & agents
Teams increasingly build software from cooperating large language model (LLM) agents, often assuming more agents mean more reliability. Yet the usual evidence, average benchmark accuracy, ignores the input mix a deployed system meets and the rule merging the agents' answers.
- 2026-11arXiv 2611.04931Modulating Merging Strengths via Joint Loss Estimation for LoRA-based Continual Learning
Kun Gu, De Cheng, Zhipeng Xu et al. · Machine learning
Continual learning (CL) aims to adapt models to new tasks while preserving previously learned knowledge. Recently, Low-Rank Adaptation (LoRA), a representative Parameter-Efficient Fine-Tuning (PEFT) method, has gained increasing attention in CL due to its efficiency and scalability.
- 2026-11arXiv 2611.02746Replay for LLM Continual Learning with Multi-Teacher On-Policy Distillation
Yiqiao Qiu · Machine learning
Multi-teacher on-policy distillation (MOPD) distills several domain experts into one student and is used in the post-training of recent large language models. It trains on all domains at once, so every expert must be available at the same time.
- 2026-11arXiv 2611.01704MatMoE: Matryoshka Mixture-of-Experts with Dynamic Mixed-Precision Quantization for Efficient Inference
Renjie Wei, Haochen Huang, Chuyu Qiu et al. · Machine learning
Mixture-of-Experts (MoE) architecture has been widely adopted in large language models (LLMs) since it reduces the computational cost through sparse expert activation.
- 2026-10arXiv 2611.00544Skill++: Towards Lifelong Agents via Routing-Aware Skill Refinement
Hanrong Zhang, Minghao Liang, Jiawei Wu et al. · AI & agents
Large language model (LLM) agents extend their capabilities with skills, each with a name, a description, and a body. Under progressive disclosure, an LLM router selects skills by name and description alone and loads only the selected bodies.
- 2026-10arXiv 2610.36994CaLLM: Continual Adaptation of LLMs via Non-Parametric Adapter Routing
Bilge Celik, Anna Vettoruzzo, Joaquin Vanschoren · Machine learning
Continual adaptation of large language models (LLMs) to evolving task distributions while mitigating catastrophic forgetting remains an open problem. Existing strategies often struggle to scale or lack a robust mechanism for zero-shot task-to-expert assignment.
- 2026-10arXiv 2610.29357Memory-Centric Continual Learning for AI Agents
Dongfang Li, Zhuoen Chen, Bonian Jia et al. · Language models & NLP
Autonomous agents powered by foundation models can reason, plan, and use tools, but their fixed weights and context limits hinder sustained learning from experience over long-term deployment.
- 2026-10arXiv 2610.25914FisherEdit: Fisher Information Guided Knowledge Editing for Large Language Models
Yizhou Yang, Jiahong Fu, Zhiwu Wang et al. · Language models & NLP
Sequential knowledge editing in Large Language Models (LLMs) often leads to a degradation of general capabilities due to cumulative parameter interference.
- 2026-10arXiv 2610.27123Reconstruction-Optimized Expert Pruning for Mixture-of-Experts Language Models
Jelena Markovic-Voronov, Petros Prastakos, Kayhan Behdin et al. · Machine learning
Mixture-of-Experts (MoE) models reduce computation through sparse expert activation, but deployment still requires storing a large number of expert parameters.
- 2026-10arXiv 2610.23213Not All Test-Time Compute Is Evidence: Model-Specific Bias in LLM Reasoning
Ming Shen, Tianjun Yao · AI & agents
Test-time scaling methods implicitly assume that additional computation from the same model provides independent evidence.
- 2026-10arXiv 2610.22293Towards Continual Learning: Mitigating Forgetting With Subspace Tuning
Arnav Chavan, Kusha Sareen, Rishabh Tiwari et al. · Machine learning
Large language models (LLMs) acquire new skills through fine-tuning, but it can destroy capabilities they already have: catastrophic forgetting remains a central obstacle to continual learning in LLMs.
- 2026-10arXiv 2610.18522Lost in Implicit Routing: How LLM Agents Fail to Load Relevant Skills at Scale
Jinlin Liu, Wei Chen, Xiaojin Zhang · AI & agents
Reusable skills let LLM agents load task-specific instructions on demand, but they make execution depend on implicit routing over skill metadata. We test whether this design scales by holding SkillsBench tasks and their golden skills fixed while adding irrelevant skills.
- 2026-10arXiv 2610.17431VeriRouter: Towards Verifiable LLM Routing
Mengyu Yao, Ziqi Zhang, Shaofei Li et al. · Security
Large language model (LLM) routers are increasingly deployed to optimize quality, latency, and cost through dynamic model selection and request rewriting.
Cited inInternal notes - 2026-10arXiv 2611.00042Diagnosing Zero-Shot Forgetting in CLIP Fine-Tuning: Attention Drift Characterizes Adaptation, Representation Drift Predicts Transfer Degradation
Ruize Xia · Vision
Fine-tuning CLIP on a downstream task typically reduces its zero-shot accuracy on other tasks, and quantifying this loss ordinarily requires a labeled transfer benchmark.
- 2026-10arXiv 2610.08791World Models' Last Exam in Physics
Mingju Gao, Qingle Liu, Yuzhao Peng et al. · Vision
Video world models can produce visually convincing yet physically inconsistent sequences, raising concerns about their reliability for prediction and planning in embodied AI systems.
- 2026-10arXiv 2610.03713What Should World Models Forget? Stratified Retention for Continual Adaptation
Nishit Anand, Ramani Duraiswami, Dinesh Manocha · Machine learning
Continual learning treats degradation on previously seen data as evidence of failure, a convention inherited from settings with a stationary prediction target, where a correct label remains correct indefinitely.
Cited inStudy: Genesys program prior art - 2026-09arXiv 2609.34453SPACE-LoRA: Allocating Activation-Subspace Protection for Continual Learning
Seunghyun Yoo, Kiseok Kim, Hyeontae Joo et al. · Machine learning
This study addresses the catastrophic forgetting problem that occurs when sequentially learning successive tasks using Low-Rank Adaptation (LoRA) from a lifelong learning perspective.
- 2026-09arXiv 2609.38224Separation of Duties for Privileged LLM Agents: A Governed Execution Architecture with Measured Security-Utility Trade-offs
Qishuai Jing · Security
Large language model agents are increasingly granted real privileges (executing commands, modifying files, calling APIs), so an agent that errs has already acted.
- 2026-09arXiv 2609.29045The Tokens Remember: When Tokenization Bypasses Knowledge Editing and Unlearning
Manit Baser, Aditya Nawal, Dinil Mon Divakaran et al. · Security
Open-weight LLMs give downstream users control over the inference stack, but this flexibility can undermine post-release guarantees that sensitive knowledge has been modified or removed.
- 2026-09arXiv 2609.27517Not What You Meant: Can LLMs Follow a Specified Negation Semantics?
Qiming Bao, Agnieszka Mensfelt, Michael J. Witbrock et al. · AI & agents
Negation does not carry a uniform interpretation across domains. In legal, regulatory, and medical reasoning, the intended interpretation depends on the reading in force -- open- versus closed-world, two- versus three-valued, and credulous versus skeptical.
- 2026-09arXiv 2609.28395Fine-Tuning LLMs for Translation: General Forgetting Mitigation Does Not Preserve MT-Specific Instruction Following
Niklas Scholz, David Thulke, Abdallah Nasir et al. · Language models & NLP
Fine-tuning large language models on parallel data improves translation quality but can cause catastrophic forgetting. Mitigation methods are generally evaluated by retention on general benchmarks.
- 2026-09arXiv 2609.25848Optimizing the Score, Losing Sight of the Task: Reward Hacking Across Weights, Selection, and Prompts
Vansh Wahi · AI & agents
A higher evaluation score does not always mean a better language model system. When optimization exploits an evaluator's mistakes, measured progress can conceal unchanged or deteriorating task performance.
Cited inInternal notes - 2026-09arXiv 2609.26035Truth for Believable AI: Expressed Doubt, Provenance, and Belief Revision as an Engineerable Stance
Sebastian Cochinescu · Language models & NLP
Conversational agents often express answers in a uniformly confident register. We test whether expressed uncertainty, provenance-aware assertion, and explicit belief revision can be implemented as a behavior layer over a fixed language model; we do not test believability or trust.
- 2026-09arXiv 2609.23916Time-Incremental Continued Pretraining of LLMs: Knowledge Updates Without Catastrophic Forgetting
Fırat Öncel, Salman Hussain Ali, Mirco Ravanelli et al. · Language models & NLP
Large language models (LLMs) drift out of date the moment their pretraining ends, yet retraining from scratch is prohibitively expensive. Continued pretraining (CPT) is the natural remedy, but it is typically evaluated through a continual learning lens that assumes disjoint data streams.
- 2026-09arXiv 2609.22055Benchmarking World Models for Continual Learning on Compositional Tasks
Haoyu Zhou, Joe Watson, Anson Lei et al. · Machine learning
A desirable property of a world model is the ability to learn continually across tasks, adapting to new environments without forgetting what the agent has already learnt.
Cited inStudy: Genesys program prior art - 2026-09arXiv 2609.19799Evolution or Illusion? Rethinking Evaluation in LLM Evolutionary Search
Tal Oved, Roi Pony, Oshri Naparstek et al. · Language models & NLP
LLM-driven evolutionary search finds programs by launching seeds and iterating each one. Papers report a single budget setting, usually one seed run for a fixed number of iterations, and rank methods from that one point.
- 2026-09arXiv 2609.17686The Missing "I Don't Know": Why Three Reasoning-Reliability Findings Converge on Calibrated Abstention
Srijith Ravikumar · Machine learning
Three recent results describe what look like unrelated LLM reliability problems. Yin et al. (2026) show reasoning RL collapses tool-reliability representations. Suleymanov et al.
- 2026-09arXiv 2609.13680Drift-Constrained Optimization: Only Direction Matters in Fine-Tuning Instruct Models
Fei Yuan, Changjiang Gao, Yilei Tu et al. · AI & agents
Fine-tuning instruct models often improves target performance while inducing behavioral drift from the reference model, which can degrade existing capabilities.
- 2026-09arXiv 2609.11108But How Would AI Agents Run a Town's Economy?
Sajal Regmi, Siddhartha Pudasaini, Chetan Phakami Pun · AI & agents
We placed 100 memory-equipped large language model (LLM) agents in charge of a closed, money-conserving spatial economy on real Pokhara Lakeside geography (earning wages, running businesses, setting prices) and ran this multi-agent simulation for up to 26 simulated weeks, well past the 1-2 weeks typical of…
- 2026-09arXiv 2609.09702Decision Shifts, Lost Label Functionality, and an Inconclusive Grounding Audit in Correctness-Gated Multi-Teacher Distillation
Xiaofei Feng · AI & agents
Candidate decision correctness and rationale grounding are different objectives. We examine correctness-gated multi-teacher distillation in a fixed experiment.
- 2026-09arXiv 2609.09776Proof-Carrying Cognition: Closing the Verification Gap with Reality-Settled Reward
Eshwar Reddy M, Sourav Karmakar · AI & agents
Frontier gains in language-model reasoning come from reinforcement learning on reasoning traces and are concentrated in domains with a cheap, sound verifier. We argue the field's binding constraint is the verification gap: no scalable, incorruptible reward for reasoning outside formal domains.
- 2026-09arXiv 2609.08067Popular Knowledge Propagates More Errors in LLM Knowledge Updating
Yuji Zhang, Weibing Wang, Cheng Qian et al. · Language models & NLP
Updating a language model's knowledge through fine-tuning is essential for keeping its outputs current, yet can also induce factual forgetting and new hallucinations.
Cited inCoherence is selected, anchored, then pruned: a literature synthesis on logical coherence in LLM world modelsEWC protects what its Fisher sees: a literature synthesis on elastic weight consolidation, scale and coherenceStudy: Elastic weight consolidation deep diveStudy: Logical coherence of LLM world models - 2026-09arXiv 2609.09113SAEScientist-Bench: Can AI Agents Conduct Autonomous SAE Interpretability Research?
Yuqiao Tan, Shizhu He, Jun Zhao et al. · AI & agents
While research on recursive self-improvement (RSI) has predominantly automated model training pipelines, reliable autonomous development demands a missing pillar: post-hoc monitoring and auditing to understand what models learn and ensure safe alignment.
- 2026-09arXiv 2609.06986Continual Learning Mechanisms Compose for Long-Horizon Memorization
Zheyuan Zhang, Alvin Zhang, Daniel Khashabi et al. · Machine learning
Language models may need to internalize information that arrives over time and retain it through many subsequent updates.
Cited inEWC protects what its Fisher sees: a literature synthesis on elastic weight consolidation, scale and coherenceTo keep what a model knows, control what it is allowed to moveStudy: Elastic weight consolidation deep diveStudy: Fortifying knowledge in continual learningStudy: Seven continual learning methods tested
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.
