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.
- 2026-07arXiv 2607.26052Spend Experts Where You Are Unsure: Confidence-Adaptive Routing for Mixture-of-Experts LoRA
Tom Saliencro, Rohan Desai, Priya Nair et al. · Machine learning
Mixture-of-Experts (MoE) variants of Low-Rank Adaptation (LoRA) route every token to a fixed number of experts . Tokens differ in how uncertain the model is about them, so a single k over-spends on easy tokens and under-serves hard ones.
- 2026-07arXiv 2607.23837Latent-LoRA: Compact Latent-Space Adapters with Gradient-Free Routing for Continual Learning
Reza Rahimi Azghan, Gautham Krishna Gudur, Giulia Pedrielli et al. · Machine learning
Large language models generalize well to individual tasks but lack an inherent mechanism for learning them sequentially, leading to catastrophic forgetting.
- 2026-07arXiv 2607.21962Ground Truth First: A Longitudinal Evaluation Instrument for Agent Memory, and the Tenure Crossover in Memory-Architecture Rankings
Quentin Spencer · Language models & NLP
Benchmarks for LLM-agent memory typically generate conversations first and extract answer keys afterwards -- with documented label-error and contamination problems -- and they overwhelmingly measure short interaction histories.
- 2026-07arXiv 2607.22157Learning on the Job: Continual Learning from Deployment Feedback for Frozen-Weights Agents
Valentin Tablan, Scott Taylor, Kristoffer Bernhem · AI & agents
AI agents encounter learning opportunities in every episode they run, and discard nearly all of them: the underlying models are frozen at deployment, so an agent that resolves a difficult request today starts from zero when it recurs tomorrow.
Cited inInternal notes - 2026-07arXiv 2607.19749The World Model Remembers, the Actor Forgets: Dream Rehearsal for Continual Model-Based RL
Gurp Nijjer · Machine learning
Model-based reinforcement-learning agents of the DreamerV3 family forget catastrophically when trained on task sequences, even when an unbounded replay buffer preserves every earlier experience.
Cited inStudy: Genesys program prior art - 2026-07arXiv 2607.17427Abliteration Is Not a Scalpel: Off-Target Effects of Refusal Removal on Decision Disposition Across Model Families
Aleksander Fafuła · Machine learning
Abliteration - deleting a model's refusal direction from its weights - is the standard recipe behind popular "uncensored" open-weight models. We show the surgery is not clean.
- 2026-07arXiv 2608.11215Poor Man's Agentic Modeling: Simulating Large LLM-Agent Societies on a Laptop
Igor Itkin · AI & agents
Simulating societies of many large language model (LLM) agents is expensive, yet the questions asked of such simulations are usually macroscopic: phase behaviour, stylised facts, and scaling with the number of agents , not the cognition of any single agent.
- 2026-07arXiv 2607.16799Diagnosing Correctness Probes under Self-Judgement Confounding
Yi-Long Lu · Language models & NLP
Hidden-state readouts can predict whether language-model outputs are correct, but objective correctness (OC) usually agrees with the model's own self-judgement (SJ), leaving the decoded signal semantically ambiguous.
- 2026-07arXiv 2607.15593Scaling LLM Agent Tool Access: The MCP Gateway
Mingxin Li, Enge Song, Yueshang Zuo et al. · Other
LLM agents increasingly rely on tool calling to complete tasks. Model Context Protocol (MCP) standardizes tool calling between LLM agents and tool providers. It reduces integration cost and drives rapid growth in tool scale.
Cited inInternal notes - 2026-07arXiv 2609.18985Suppressed, Not Erased: A Representational Trace of Edited Facts Survives Even Weight-Free Knowledge Editing
Priyansh Srivastava, Romit Chatterjee · AI & agents
Knowledge-editing benchmarks certify local correctness, whether an edited model produces the new fact on near-edit prompts but not how much of the original fact remains decodable inside the model.
- 2026-07arXiv 2607.14528Controlled Reformulation Testing for Logical Consistency in Large Language Models
Alexander Gu, Alan Chen · Language models & NLP
Large language models (LLMs) frequently contradict themselves when the surface form of a logically equivalent question changes. We present a benchmark of 350 question families (1,750 total questions) for Controlled Reformulation Testing (CRTBench) to evaluate logical invariance.
- 2026-07arXiv 2607.13568Graded Entity-Familiarity Readouts in Language Models: Polish Adaptation, Cross-Language Robustness, and Refusal Steering
Grzegorz Brzezinka · Language models & NLP
Can a language model estimate its familiarity with an entity before generating an answer?
- 2026-07arXiv 2607.14004Do Agent Optimizers Compound? A Continual-Learning Evaluation on Terminal-Bench 2.0
Wenxiao Wang, Priyatham Kattakinda, Soheil Feizi · AI & agents
Most reported gains from agent-optimization methods are one-shot: an agent is optimized against a fixed benchmark and the resulting improvement is reported as if it were a stable property of the method.
Cited inInternal notes - 2026-07arXiv 2607.11020Can a Language Model Learn Facts Continually in Its Weights?
Charles O'Neill, Max Kirkby, Jonathon Liu et al. · Language models & NLP
Continual learning is a long-standing capability gap between LLMs and humans. Writing new knowledge into a model's weights routinely causes it to forget old knowledge, commonly denoted as "catastrophic forgetting".
- 2026-07arXiv 2607.11997Are we Merging the Right Models? Impact of Expert Training Duration on Model Merging for LLMs
Nikita Kozodoi, Zainab Afolabi, Jack Butler · Machine learning
Multi-task model merging combines separately trained expert models into a single model that handles all tasks without co-training. Standard practice merges experts at their optimal validation loss.
- 2026-07arXiv 2607.10738To Answer or to Abstain: Mitigating Search-Agent Hallucinations via Abstention-Aware Reinforcement Learning
Fengji Zhang, Tianyu Fan, Yuxiang Zheng et al. · Machine learning
Recent advances in equipping Large Language Models (LLMs) with search tools and outcome-reward reinforcement learning (RL) have achieved new state-of-the-art results on open-domain QA tasks.
- 2026-07arXiv 2607.10113Dynamic Agent Skills: A Lifecycle Survey and Taxonomy of Evolving Skill Libraries
Yubo Li · Transactions on Machine Learning Research, 2026 · AI & agents
Large language model agents increasingly store reusable procedures outside the model.md packages, workflow graphs, or learned adapters that a future agent can retrieve and invoke.
Cited inInternal notes - 2026-07arXiv 2607.10362A Control Theory of Predictability in Latent World Models
Hanzhe You, Yonggang Zhang, Maohao Ran et al. · Machine learning
Latent world models are trained to predict future states in a learned representation and are then deployed inside a planner that selects actions by simulating them forward.
- 2026-07arXiv 2607.09015Correlation-Aware Contextual Bandits with Surrogate Rewards for LLM Routing
Ajay Narayanan Sridhar, Ronak Singh, Mehrdad Mahdavi et al. · Machine learning
We study contextual bandit problems with correlated arms and access to surrogate reward signals produced by a machine learning model, motivated by applications such as large language model (LLM) routing.
- 2026-07arXiv 2607.08065When LLMs Agree, Are They Right? Auditing Self-Consistency and Cross-Model Agreement as Confidence Signals
Kaihua Ding · AI & agents
, 2017) panels of judges. These systems share a key assumption: that consistency -- agreement among judges, or among a model's own samples -- indicates correctness.
- 2026-07arXiv 2607.08393Towards Mechanistically Understanding Why Memorized Knowledge Fails to Generalize in Large Language Model Finetuning
Lu Dai, Ziyang Rao, Yili Wang et al. · AI & agents
Fine-tuning LLMs to inject new knowledge faces a critical challenge: LLMs can quickly memorize new facts, yet fail to use them for downstream reasoning tasks.
- 2026-07arXiv 2608.14611The 2026 Singapore Consensus on Global AI Safety Research Priorities
Stephen Casper, Oskar Galeev, Yoshua Bengio et al. · Society, software & economics
Frontier AI capabilities and autonomy are advancing rapidly. A growing number of real-world incidents make a trusted AI ecosystem essential to embracing AI with confidence.
- 2026-07arXiv 2607.07050When Top-K Misses the Decision: Tool-Call Drift in Multi-Teacher On-Policy Distillation
Jiabin Shen, Guang Chen, Chengjun Mao · Language models & NLP
Top- teacher logits make on-policy distillation tractable, but retained teacher mass does not certify student-relative gradient fidelity. We study routed, two-teacher tool-use distillation.
- 2026-07arXiv 2607.05904More Convincing, Not More Correct: Self-Play Reward Hacking of Reference-Free LLM Judges
Chenyu Zhou · Machine learning
Training a language model against its own reference-free judgments (the premise of self-rewarding, self-play, and LLM-as-a-judge pipelines) assumes a model's verdict on a shown answer tracks correctness.
Cited inInternal notes - 2026-07arXiv 2607.06202UBEP: Re-architecting Expert Parallelism Communication Library for Production Superpods
Yipeng Liu, Chang Liu, Si Shen et al. · Other
The deployment of Mixture-of-Experts (MoE) models on production high-bandwidth superpods, such as NVIDIA's NVL72/576 and Huawei's CloudMatrix384, introduces critical challenges beyond raw interconnect bandwidth.
Cited inThe Wrong Phase - 2026-07arXiv 2607.06401A Definition and Roadmap for World Models
Xinyuan Chen, Haoyu Guo, Shi Guo et al. · AI & agents
World models -- internal simulators that learn the structure and dynamics of an environment -- have become one of the most actively debated concepts in AI.
- 2026-07arXiv 2607.04364RL Forgets! Towards Continual Policy Optimization
Mao-Lin Luo, Zhe-Xu Wang, Zi-Hao Zhou et al. · Machine learning
Continual post-training is becoming a central paradigm for adapting vision-language models to evolving tasks. Recent studies report that reinforcement learning is less prone to forgetting than supervised fine-tuning, motivating the view that RL is inherently resistant to forgetting.
- 2026-07arXiv 2607.02964Individual Parameters in Weight-Sparse Transformers Appear Interpretable
Arnau Marin-Llobet, Stefan Heimersheim · Machine learning
A central goal of mechanistic interpretability is to understand how neural networks work and what each individual component does. Dominant circuit-finding approaches focus on a specific behavior and reverse-engineer the role of components on the associated sub-distribution.
- 2026-07arXiv 2607.01763Denser Better: Limits of On-Policy Self-Distillation for Continual Post-Training
Meng Wang, Haohan Zhao, Wenzhuo Liu et al. · Machine learning
Continual post-training enables foundation models to acquire new knowledge while preserving existing capabilities. Recent work suggests that on-policy learning can mitigate forgetting, with self-distillation as a particularly attractive approach.
- 2026-07arXiv 2607.01874SkillCoach: Self-Evolving Rubrics for Evaluating and Enhancing Agentic Skill-Use
Jiayin Zhu, Kelong Mao, Yudong Guo et al. · AI & agents
Skills are becoming a reusable operational layer for LLM agents, encoding SOPs, domain rules, tool workflows, scripts, and validation routines. In realistic skill repositories, overlapping skills make reliable skill-use difficult.
Cited inInternal notes - 2026-07arXiv 2608.07514Open Technical Problems in Open-Weight AI Model Risk Management
Stephen Casper, Kyle O'Brien, Shayne Longpre et al. · Society, software & economics
Frontier AI models with openly available weights are steadily becoming more powerful and widely adopted. However, compared to proprietary models, open-weight models pose different opportunities and challenges for effective risk management.
- 2026-06arXiv 2606.31174ClawArena-Team: Benchmarking Subagent Orchestration and Dynamic Workflows in Language-Model Agents
Kaiwen Xiong, Haonian Ji, Shi Qiu et al. · AI & agents
Production large language-model (LLM) agents are increasingly deployed not as lone problem-solvers but as managers: a main model creates specialized subagents, delegates work, and orchestrates their parallel, asynchronous returns through dynamic workflows.
- 2026-06arXiv 2606.31404Wisdom Of The (AI) Crowd: Investigating Artificial Swarm Intelligence In Large Language Models
Justin Brenne, Christian Meske · AI & agents
Human swarm intelligence demonstrates remarkable collective accuracy but faces scalability constraints in cost, coordination, and time.
- 2026-06arXiv 2606.30306Always-OnAgents:A Survey of Persistent Memory, State, and Governance in LLMAgents
Tianyu Ding, Aditya Nannapaneni, Bingfan Liu et al. · AI & agents
Always-on agents are systems whose future behavior depends on durable state accumulated across earlier interactions.
Cited inInternal notes - 2026-06arXiv 2606.30317MCP Server Architecture Patterns for LLM-Integrated Applications
Carson Rodrigues, Oysturn Vas · Society, software & economics
The Model Context Protocol (MCP), introduced by Anthropic in November 2024, defines a standardized interface for connecting large language models (LLMs) to external tools, data sources, and services.
- 2026-06arXiv 2606.29270Minority Sentinel: When to Overturn Majority Voting in Multi-Agent LLM Debates
Chuan He, Zebin Chen, Zhengyi Yang et al. · AI & agents
Multi-Agent Debate (MAD) with Majority Voting is a dominant paradigm for improving LLM reasoning, yet its effectiveness rests on the Condorcet Jury Theorem's assumption of independent errors.
- 2026-06arXiv 2606.28757A Physics-Grounded Benchmark for Multi-Agent Dynamics in World Models
Nuo Chen, Lulin Liu, Zihao Li et al. · Vision
Generative world models hold immense promise as scalable simulators for autonomous systems, particularly for synthesizing rare but safety-critical multi-agent interactions, such as vehicle collisions.
- 2026-06arXiv 2606.26783Reproducibility Study of "AlphaEdit: Null-Space Constrained Knowledge Editing for Language Models"
Ananth K Suresh, Arya Hariharan · Machine learning
Fang et al.
- 2026-06arXiv 2606.27326Hallucination in World Models is Predictable and Preventable
Nicklas Hansen, Xiaolong Wang · Machine learning
Modern generative world models render increasingly realistic action-controllable futures, yet they frequently hallucinate: rollouts remain visually fluent while drifting from the ground-truth dynamics.
- 2026-06arXiv 2606.27374World Action Models Enable Continual Imitation Learning with Recurrent Generative Replays
Manish Kumar Govind, Dominick Reilly, Smit Patel et al. · Robotics & control
Going beyond predicting robot actions, World Action Models (WAMs) can also generate future visual observations.
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.
