PLD+: Accelerating LLM inference by leveraging Language Model Artifacts

To reduce the latency associated with autoretrogressive LLM inference, speculative decoding has emerged as a novel decoding paradigm, where future tokens are drafted and verified in parallel. However, the practical deployment of speculative decoding is hindered by its requirements for additional computational resources and fine-tuning, which limits its out-of-the-box usability. To address these challenges, we present PLD+, a suite of novel algorithms developed to accelerate the inference process of LLMs, particularly for input-guided tasks. These tasks, which include code editing, text editing, summarization, etc., often feature outputs with substantial overlap with their inputs-an attribute PLD+ is designed to exploit. PLD+ also leverages the artifacts (attention and hidden states) generated during inference to accelerate inference speed. We test our approach on five input-guided tasks and through extensive experiments we find that PLD+ outperforms all tuning-free approaches. In the greedy setting, it even outperforms the state-of-the-art tuning-dependent approach EAGLE on four of the tasks. (by a margin of upto 2.31 in terms of avg. speedup). Our approach is tuning free, does not require any additional compute and can easily be used for accelerating inference of any LLM.

Beyond Logit Lens: Contextual Embeddings for Robust Hallucination Detection & Grounding in VLMs

The rapid development of Large Multimodal Models (LMMs) has significantly advanced multimodal understanding by harnessing the language abilities of Large Language Models (LLMs) and integrating modality-specific encoders. However, LMMs are plagued by hallucinations that limit their reliability and adoption. While traditional methods to detect and mitigate these hallucinations often involve costly training or rely heavily on external models, recent approaches utilizing internal model features present a promising alternative. In this paper, we critically assess the limitations of the state-of-the-art training-free technique, the logit lens, in handling generalized visual hallucinations. We introduce a refined method that leverages contextual token embeddings from middle layers of LMMs. This approach significantly improves hallucination detection and grounding across diverse categories, including actions and OCR, while also excelling in tasks requiring contextual understanding, such as spatial relations and attribute comparison. Our novel grounding technique yields highly precise bounding boxes, facilitating a transition from Zero-Shot Object Segmentation to Grounded Visual Question Answering. Our contributions pave the way for more reliable and interpretable multimodal models.

Enhancing Post-Hoc Attributions in Long Document Comprehension via Coarse Grained Answer Decomposition

Accurately attributing answer text to its source document is crucial for developing a reliable question-answering system. However, attribution for long documents remains largely unexplored. Post-hoc attribution systems are designed to map answer text back to the source document, yet the granularity of this mapping has not been addressed. Furthermore, a critical question arises: What precisely should be attributed, with an emphasis on identifying the information units within an answer that necessitate grounding? In this paper, we propose and investigate a novel approach to the factual decomposition of generated answers for attribution, employing template-based in-context learning. To accomplish this, we utilize the question and integrate negative sampling during few-shot in-context learning for decomposition. This approach enhances the semantic understanding of both abstractive and extractive answers. We examine the impact of answer decomposition by providing a thorough examination of various attribution approaches, ranging from retrieval-based techniques to LLM-based attributors.

Enhancing Presentation Slide Generation by LLMs with a Multi-Staged End-to-End Approach

Generating presentation slides from a long document with multimodal elements such as text and images is an important task. This is time consuming and needs domain expertise if done manually. Existing approaches for generating a rich presentation from a document are often semi-automatic or only put a flat summary into the slides ignoring the importance of a good narrative. In this paper, we address this research gap by proposing a multi-staged end-to-end model which uses a combination of LLM and VLM. We have experimentally shown that compared to applying LLMs directly with state-of-the-art prompting, our proposed multi-staged solution is better in terms of automated metrics and human evaluation.

Peering into the Mind of Language Models: An Approach for Attribution in Contextual Question Answering

With the enhancement in the field of generative artificial intelligence (AI), contextual question answering has become extremely relevant. Attributing model generations to the input source document is essential to ensure trustworthiness and reliability. We observe that when large language models (LLMs) are used for contextual question answering, the output answer often consists of text copied verbatim from the input prompt which is linked together with "glue text" generated by the LLM. Motivated by this, we propose that LLMs have an inherent awareness from where the text was copied, likely captured in the hidden states of the LLM. We introduce a novel method for attribution in contextual question answering, leveraging the hidden state representations of LLMs. Our approach bypasses the need for extensive model retraining and retrieval model overhead, offering granular attributions and preserving the quality of generated answers. Our experimental results demonstrate that our method performs on par or better than GPT-4 at identifying verbatim copied segments in LLM generations and in attributing these segments to their source. Importantly, our method shows robust performance across various LLM architectures, highlighting its broad applicability. Additionally, we present Verifiability-granular, an attribution dataset which has token level annotations for LLM generations in the contextual question answering setup.

Towards Optimizing the Costs of LLM Usage

Generative AI and LLMs in particular are heavily used nowadays for various document processing tasks such as question answering and summarization. However, different LLMs come with different capabilities for different tasks as well as with different costs, tokenization, and latency. In fact, enterprises are already incurring huge costs of operating or using LLMs for their respective use cases. In this work, we propose optimizing the usage costs of LLMs by estimating their output quality (without actually invoking the LLMs), and then solving an optimization routine for the LLM selection to either keep costs under a budget, or minimize the costs, in a quality and latency aware manner. We propose a model to predict the output quality of LLMs on document processing tasks like summarization, followed by an LP rounding algorithm to optimize the selection of LLMs. We study optimization problems trading off the quality and costs, both theoretically and empirically. We further propose a sentence simplification model for reducing the number of tokens in a controlled manner. Additionally, we propose several deterministic heuristics for reducing tokens in a quality aware manner, and study the related optimization problem of applying the heuristics optimizing the quality and cost trade-off. We perform extensive empirical validation of our methods on not only enterprise datasets but also on open-source datasets, annotated by us, and show that we perform much better compared to closest baselines. Our methods reduce costs by 40%- 90% while improving quality by 4%-7%. We will release the annotated open source datasets to the community for further research and exploration.

Social Media Ready Caption Generation for Brands

Social media advertisements are key for brand marketing, aiming to attract consumers with captivating captions and pictures or logos. While previous research has focused on generating captions for general images, incorporating brand personalities into social media captioning remains unexplored. Brand personalities are shown to be affecting consumers' behaviours and social interactions and thus are proven to be a key aspect of marketing strategies. Current open-source multimodal LLMs are not directly suited for this task. Hence, we propose a pipeline solution to assist brands in creating engaging social media captions that align with the image and the brand personalities. Our architecture is based on two parts: a the first part contains an image captioning model that takes in an image that the brand wants to post online and gives a plain English caption; b the second part takes in the generated caption along with the target brand personality and outputs a catchy personality-aligned social media caption. Along with brand personality, our system also gives users the flexibility to provide hashtags, Instagram handles, URLs, and named entities they want the caption to contain, making the captions more semantically related to the social media handles. Comparative evaluations against various baselines demonstrate the effectiveness of our approach, both qualitatively and quantitatively.

Drilling Down into the Discourse Structure with LLMs for Long Document Question Answering

We address the task of evidence retrieval for long document question answering, which involves locating relevant paragraphs within a document to answer a question. We aim to assess the applicability of large language models (LLMs) in the task of zero-shot long document evidence retrieval, owing to their unprecedented performance across various NLP tasks. However, currently the LLMs can consume limited context lengths as input, thus providing document chunks as inputs might overlook the global context while missing out on capturing the inter-segment dependencies. Moreover, directly feeding the large input sets can incur significant computational costs, particularly when processing the entire document (and potentially incurring monetary expenses with enterprise APIs like OpenAI's GPT variants). To address these challenges, we propose a suite of techniques that exploit the discourse structure commonly found in documents. By utilizing this structure, we create a condensed representation of the document, enabling a more comprehensive understanding and analysis of relationships between different parts. We retain $99.6\%$ of the best zero-shot approach's performance, while processing only $26\%$ of the total tokens used by the best approach in the information seeking evidence retrieval setup. We also show how our approach can be combined with \textit{self-ask} reasoning agent to achieve best zero-shot performance in complex multi-hop question answering, just $\approx 4\%$ short of zero-shot performance using gold evidence.

A-STAR: Test-time Attention Segregation and Retention for Text-to-image Synthesis

While recent developments in text-to-image generative models have led to a suite of high-performing methods capable of producing creative imagery from free-form text, there are several limitations. By analyzing the cross-attention representations of these models, we notice two key issues. First, for text prompts that contain multiple concepts, there is a significant amount of pixel-space overlap (i.e., same spatial regions) among pairs of different concepts. This eventually leads to the model being unable to distinguish between the two concepts and one of them being ignored in the final generation. Next, while these models attempt to capture all such concepts during the beginning of denoising (e.g., first few steps) as evidenced by cross-attention maps, this knowledge is not retained by the end of denoising (e.g., last few steps). Such loss of knowledge eventually leads to inaccurate generation outputs. To address these issues, our key innovations include two test-time attention-based loss functions that substantially improve the performance of pretrained baseline text-to-image diffusion models. First, our attention segregation loss reduces the cross-attention overlap between attention maps of different concepts in the text prompt, thereby reducing the confusion/conflict among various concepts and the eventual capture of all concepts in the generated output. Next, our attention retention loss explicitly forces text-to-image diffusion models to retain cross-attention information for all concepts across all denoising time steps, thereby leading to reduced information loss and the preservation of all concepts in the generated output.

Friendly Neighbors: Contextualized Sequence-to-Sequence Link Prediction

We propose KGT5-context, a simple sequence-to-sequence model for link prediction (LP) in knowledge graphs (KG). Our work expands on KGT5, a recent LP model that exploits textual features of the KG, has small model size, and is scalable. To reach good predictive performance, however, KGT5 relies on an ensemble with a knowledge graph embedding model, which itself is excessively large and costly to use. In this short paper, we show empirically that adding contextual information - i.e., information about the direct neighborhood of a query vertex - alleviates the need for a separate KGE model to obtain good performance. The resulting KGT5-context model obtains state-of-the-art performance in our experimental study, while at the same time reducing model size significantly.