arXiv subjects

Philip Cohen

Publications and source records attributed to Philip Cohen.

5 recordsLinked to original sources

arXiv2024

Improving Cross-Domain Low-Resource Text Generation through LLM Post-Editing: A Programmer-Interpreter Approach

Post-editing has proven effective in improving the quality of text generated by large language models (LLMs) such as GPT-3.5 or GPT-4, particularly when direct updating of their parameters to enhance text quality is infeasible or expensive. However, relying solely on smaller language models for post-editing can limit the LLMs' ability to generalize across domains. Moreover, the editing strategies in these methods are not optimally designed for text-generation tasks. To address these limitations, we propose a neural programmer-interpreter approach that preserves the domain generalization ability of LLMs when editing their output. The editing actions in this framework are specifically devised for text generation. Extensive experiments demonstrate that the programmer-interpreter significantly enhances GPT-3.5's performance in logical form-to-text conversion and low-resource machine translation, surpassing other state-of-the-art (SOTA) LLM post-editing methods in cross-domain settings.

cs.CL↗

arXiv2023

Reranking for Natural Language Generation from Logical Forms: A Study based on Large Language Models

Large language models (LLMs) have demonstrated impressive capabilities in natural language generation. However, their output quality can be inconsistent, posing challenges for generating natural language from logical forms (LFs). This task requires the generated outputs to embody the exact semantics of LFs, without missing any LF semantics or creating any hallucinations. In this work, we tackle this issue by proposing a novel generate-and-rerank approach. Our approach involves initially generating a set of candidate outputs by prompting an LLM and subsequently reranking them using a task-specific reranker model. In addition, we curate a manually collected dataset to evaluate the alignment between different ranking metrics and human judgements. The chosen ranking metrics are utilized to enhance the training and evaluation of the reranker model. By conducting extensive experiments on three diverse datasets, we demonstrate that the candidates selected by our reranker outperform those selected by baseline methods in terms of semantic consistency and fluency, as measured by three comprehensive metrics. Our findings provide strong evidence for the effectiveness of our approach in improving the quality of generated outputs.

cs.CL↗

arXiv2021

Moments of discrete classical $q$-orthogonal polynomial ensembles

We consider some discrete $q$-analogues of the classical continuous orthogonal polynomial ensembles. Building on results due to Morozov, Popolitov and Shakirov, we find representations for the moments of the discrete $q$-Hermite and discrete $q$-Laguerre ensembles in terms of basic hypergeometric series. We find that when the number of particles is suitably randomised, the moments may be represented as basic hypergeometric orthogonal polynomials, with corresponding three-term recurrences in $k$, the order of the moments.

math.PR↗

arXiv2020

Moments of discrete orthogonal polynomial ensembles

We obtain factorial moment identities for the Charlier, Meixner and Krawtchouk orthogonal polynomial ensembles. Building on earlier results by Ledoux [Elect. J. Probab. 10, (2005)], we find hypergeometric representations for the factorial moments when the reference measure is Poisson (Charlier ensemble) and geometric (a particular case of the Meixner ensemble). In these cases, if the number of particles is suitably randomised, the factorial moments have a polynomial property, and satisfy three-term recurrence relations and differential equations. In particular, the normalised factorial moments of the randomised ensembles are precisely related to the moments of the corresponding equilibrium measures. We also briefly outline how these results can be interpreted as Cauchy-type identities for certain Schur measures.

math.PR↗

arXiv2014

Wide-Gap Semiconducting Graphene from Nitrogen-Seeded SiC

All carbon electronics based on graphene has been an elusive goal. For more than a decade, the inability to produce significant band-gaps in this material has prevented the development of semiconducting graphene. While chemical functionalization was thought to be a route to semiconducting graphene, disorder in the chemical adsorbates, leading to low mobilities, have proved to be a hurdle in its production. We demonstrate a new approach to produce semiconducting graphene that uses a small concentration of covalently bonded surface nitrogen, not as a means to functionalize graphene, but instead as a way to constrain and bend graphene. We demonstrate that a submonolayer concentration of nitrogen on SiC is sufficient to pin epitaxial graphene to the SiC interface as it grows, causing the graphene to buckle. The resulting 3-dimensional modulation of the graphene opens a band-gap greater than 0.7eV in the otherwise continuous metallic graphene sheet.

cond-mat.mtrl-sci↗