# Taming LLMs with Negative Samples: A Reference-Free Framework to Evaluate Presentation Content with Actionable Feedback

arXiv: https://arxiv.org/abs/2505.18240 · HTML: https://mlsanity.com/paper/2505.18240/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 25 of its 37 references that are in the map.
- First posted on arXiv 23 May 2025.

## Builds on (most cited first, 25 of 25)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks](https://mlsanity.com/paper/1908.10084/) (2019)
- [BERTScore: Evaluating Text Generation with BERT](https://mlsanity.com/paper/1904.09675/) (2019)
- [G-Eval: NLG Evaluation using GPT-4 with Better Human Alignment](https://mlsanity.com/paper/2303.16634/) (2023)
- [MoverScore: Text Generation Evaluating with Contextualized Embeddings and Earth Mover Distance](https://mlsanity.com/paper/1909.02622/) (2019)
- [Towards Question-Answering as an Automatic Metric for Evaluating the Content Quality of a Summary](https://mlsanity.com/paper/2010.00490/) (2020)
- [Hard negative examples are hard, but useful](https://mlsanity.com/paper/2007.12749/) (2020)
- [Are Large Language Models Capable of Generating Human-Level Narratives?](https://mlsanity.com/paper/2407.13248/) (2024)
- [ChatGPT as a Factual Inconsistency Evaluator for Text Summarization](https://mlsanity.com/paper/2303.15621/) (2023)
- [Efficient multi-prompt evaluation of LLMs](https://mlsanity.com/paper/2405.17202/) (2024)
- [SUPERT: Towards New Frontiers in Unsupervised Evaluation Metrics for Multi-Document Summarization](https://mlsanity.com/paper/2005.03724/) (2020)
- [On the Limitations of Reference-Free Evaluations of Generated Text](https://mlsanity.com/paper/2210.12563/) (2022)
- [D2S: Document-to-Slide Generation Via Query-Based Text Summarization](https://mlsanity.com/paper/2105.03664/) (2021)
- [Finding Blind Spots in Evaluator LLMs with Interpretable Checklists](https://mlsanity.com/paper/2406.13439/) (2024)
- [FENICE: Factuality Evaluation of summarization based on Natural language Inference and Claim Extraction](https://mlsanity.com/paper/2403.02270/) (2024)
- [A Semantic QA-Based Approach for Text Summarization Evaluation](https://mlsanity.com/paper/1704.06259/) (2017)
- [Enhancing Presentation Slide Generation by LLMs with a Multi-Staged End-to-End Approach](https://mlsanity.com/paper/2406.06556/) (2024)
- [A Comparative Study of Quality Evaluation Methods for Text Summarization](https://mlsanity.com/paper/2407.00747/) (2024)
- [Presentations are not always linear! GNN meets LLM for Document-to-Presentation Transformation with Attribution](https://mlsanity.com/paper/2405.13095/) (2024)
- [HAUSER: Towards Holistic and Automatic Evaluation of Simile Generation](https://mlsanity.com/paper/2306.07554/) (2023)
- [Unveiling Narrative Reasoning Limits of Large Language Models with Trope in Movie Synopses](https://mlsanity.com/paper/2409.14324/) (2024)
- [EasyJudge: an Easy-to-use Tool for Comprehensive Response Evaluation of LLMs](https://mlsanity.com/paper/2410.09775/) (2024)
- [Reading Order Independent Metrics for Information Extraction in Handwritten Documents](https://mlsanity.com/paper/2404.18664/) (2024)

## Built on by (most cited first, 2 of 2)

- [DeepSlide: From Artifacts to Presentation Delivery](https://mlsanity.com/paper/2605.15202/) (2026)
- [SLIDEFORGE: An LLM Agent for Controllable Editing of Slides as Structured Artifacts](https://mlsanity.com/paper/2609.03109/) (2026)

## Abstract (arXiv)

The generation of presentation slides automatically is an important problem in the era of generative AI. This paper focuses on evaluating multimodal content in presentation slides that can effectively summarize a document and convey concepts to a broad audience. We introduce a benchmark dataset, RefSlides, consisting of human-made high-quality presentations that span various topics. Next, we propose a set of metrics to characterize different intrinsic properties of the content of a presentation and present REFLEX, an evaluation approach that generates scores and actionable feedback for these metrics. We achieve this by generating negative presentation samples with different degrees of metric-specific perturbations and use them to fine-tune LLMs. This reference-free evaluation technique does not require ground truth presentations during inference. Our extensive automated and human experiments demonstrate that our evaluation approach outperforms classical heuristic-based and state-of-the-art large language model-based evaluations in generating scores and explanations.