Papers where the original authors have engaged with our plain-language explanation.
For every paper on this page, at least one of the original authors has seen our plain-language explanation and engaged with it — either confirming it reads accurately or requesting corrections that we then applied. An endorsement does not mean the authors formally approve every sentence, but it does mean the explanation has passed the eyes of the people who wrote the paper.
438 papers reviewed by authors · 281–290 / 438
This paper introduces "Federation over Text" (FoT), a gradient-free framework where multiple LLM agents iteratively share and aggregate their reasoning traces into a central metacognitive insight library, significantly improving reasoning accuracy and efficiency across diverse tasks without requiring gradient optimization or supervision.
This paper investigates the potential and limitations of large language models in medical decision-making by distinguishing between the "treatment problem" (requiring causal inference from evidence) and the "chat problem" (relying on imitation), while using statins as a case study to highlight the ethical and observational challenges in training AI for clinical care.
This paper proposes an asymmetrical two-brane model where the Higgs vacuum expectation value varies between branes, resulting in superheavy fermions on a second, observer-free brane that could constitute a component of dark matter and generate ultra-high-energy particles through inter-brane photon interactions.
This paper proposes a theoretical framework for finite algorithmics and a generic method for automatically discovering efficient solutions to NP-complete problems in practice, arguing that bounded input sizes make these problems easier to solve than their general asymptotic counterparts.
This paper introduces ClawXiv, a local-first, signed archival workflow and distributed publication architecture designed to transform volatile human-AI collaborative research sessions and heterogeneous files into durable, inspectable, and content-addressed artifacts through a four-state lifecycle.
This paper presents a comparative study of Integrated Gradients, Attention Rollout, and SHAP on a fine-tuned DistilBERT model for sentiment classification, revealing that gradient-based methods offer more stable and intuitive explanations while highlighting the trade-offs between computational efficiency, flexibility, and reliability in practical LLM explainability.
This study shows that epitaxial HfZrO-based ferroelectric memristive devices exhibit coexisting area-dependent tunneling and localized-conduction regimes, with a statistical crossover at approximately that correlates with the onset of ferroelectric wake-up and oxygen-vacancy redistribution.
This study utilizes deep VLT/HAWK-I photometry to measure the disk fraction in the high-mass star-forming region M17, finding that while local UV flux shows no correlation with disk survival due to dynamical mixing, a comparison with other regions of similar age confirms that external photoevaporation significantly reduces average disk lifetimes.
This paper introduces -FlowNet, a deep learning framework utilizing an attention-based U-Net architecture to efficiently and accurately predict fluid flow patterns in irregular microchannels, outperforming standard U-Net and T-Net models with superior dice scores and intersection over union metrics while overcoming the computational limitations of traditional CFD simulations.
The paper demonstrates that a rational agent possessing true and refinable knowledge of events can never determine whether their knowledge is complete, a limitation that persists regardless of introspection or the acquisition of new information.