Large language models (LLMs) have emerged and presented their general problem-solving capabilities with one model. However, the model size has increased dramatically with billions of parameters to enable such broad problem-solving capabilities. In addition, due to the dominance of matrix-matrix and matrix-vector multiplications in LLMs, the compute-to-model size ratio is significantly lower than that of CNNs. This shift pushes LLMs from a computation-bound regime to a memory-bound regime. Therefore, optimizing the memory footprint and traffic is an important optimization direction for LLMs today. Model compression methods such as quantization and parameter pruning have been actively explored for achieving the memory footprint and traffic optimization. However, the accuracy-efficiency trade-off of rank pruning for LLMs is not well-understood yet. Therefore, we characterize the accuracy-efficiency trade-off of a low-rank decomposition method, specifically Tucker decomposition, on recent language models, including an open-source LLM, Llama 2. We formalize the low-rank decomposition design space and show that the decomposition design space is enormous (e.g., O($2^{37}$) for Llama2-7B). To navigate such a vast design space, we formulate the design space and perform thorough case studies of accuracy-efficiency trade-offs using six widely used LLM benchmarks on BERT and Llama 2 models. Our results show that we can achieve a 9\% model size reduction with minimal accuracy drops, which range from 4\%p to 10\%p, depending on the difficulty of the benchmark, without any retraining to recover accuracy after decomposition. The results show that low-rank decomposition can be a promising direction for LLM-based applications that require real-time service in scale (e.g., AI agent assist and real-time coding assistant), where the latency is as important as the model accuracy.

The trustworthiness of highly capable language models is put at risk when they are able to produce deceptive outputs. Moreover, when models are vulnerable to deception it undermines reliability. In this paper, we introduce a method to investigate complex, model-on-model deceptive scenarios. We create a dataset of over 10,000 misleading explanations by asking Llama-2 7B, 13B, 70B, and GPT-3.5 to justify the wrong answer for questions in the MMLU. We find that, when models read these explanations, they are all significantly deceived. Worryingly, models of all capabilities are successful at misleading others, while more capable models are only slightly better at resisting deception. We recommend the development of techniques to detect and defend against deception. Since the release of OpenAI's ChatGPT, large language models (LLMs) have revolutionized information accessibility by providing precise answers and supportive explanations to complex queries (Spatharioti et al., 2023; Caramancion, 2024; OpenAI, 2022). However, LLMs have also demonstrated a propensity to hallucinate explanations that are convincing but incorrect (Zhang et al., 2023; Walters & Wilder, 2023; Xu et al., 2024).

Unified multi-model representation spaces are the foundation of multimodal understanding and generation. However, the billions of model parameters and catastrophic forgetting problems make it challenging to further enhance pre-trained unified spaces. In this work, we propose FreeBind, an idea that treats multimodal representation spaces as basic units, and freely augments pre-trained unified space by integrating knowledge from extra expert spaces via "space bonds". Specifically, we introduce two kinds of basic space bonds: 1) Space Displacement Bond and 2) Space Combination Bond. Based on these basic bonds, we design Complex Sequential & Parallel Bonds to effectively integrate multiple spaces simultaneously. Benefiting from the modularization concept, we further propose a coarse-to-fine customized inference strategy to flexibly adjust the enhanced unified space for different purposes. Experimentally, we bind ImageBind with extra image-text and audio-text expert spaces, resulting in three main variants: ImageBind++, InternVL_IB, and InternVL_IB++. These resulting spaces outperform ImageBind on 5 audio-image-text downstream tasks across 9 datasets. Moreover, via customized inference, it even surpasses the advanced audio-text and image-text expert spaces.

Federated Learning faces significant challenges in statistical and system heterogeneity, along with high energy consumption, necessitating efficient client selection strategies. Traditional approaches, including heuristic and learning-based methods, fall short of addressing these complexities holistically. In response, we propose FedGCS, a novel generative client selection framework that innovatively recasts the client selection process as a generative task. Drawing inspiration from the methodologies used in large language models, FedGCS efficiently encodes abundant decision-making knowledge within a continuous representation space, enabling efficient gradient-based optimization to search for optimal client selection that will be finally output via generation. The framework comprises four steps: (1) automatic collection of diverse "selection-score" pair data using classical client selection methods; (2) training an encoder-evaluator-decoder framework on this data to construct a continuous representation space; (3) employing gradient-based optimization in this space for optimal client selection; (4) generating the final optimal client selection via using beam search for the well-trained decoder. FedGCS outperforms traditional methods by being more comprehensive, generalizable, and efficient, simultaneously optimizing for model performance, latency, and energy consumption. The effectiveness of FedGCS is proven through extensive experimental analyses.

Given recent advancement of Large Language Models (LLMs), the task of translating from natural language prompts to different programming languages (code generation) attracts immense attention for wide application in different domains. Specially code generation for Bash (NL2Bash) is widely used to generate Bash scripts for automating different tasks, such as performance monitoring, compilation, system administration, system diagnostics, etc. Besides code generation, validating synthetic code is critical before using them for any application. Different methods for code validation are proposed, both direct (execution evaluation) and indirect validations (i.e. exact/partial match, BLEU score). Among these, Execution-based Evaluation (EE) can validate the predicted code by comparing the execution output of model prediction and expected output in system. However, designing and implementing such an execution-based evaluation system for NL2Bash is not a trivial task. In this paper, we present a machinery for execution-based evaluation for NL2Bash. We create a set of 50 prompts to evaluate some popular LLMs for NL2Bash. We also analyze several advantages and challenges of EE such as syntactically different yet semantically equivalent Bash scripts generated by different LLMs, or syntactically correct but semantically incorrect Bash scripts, and how we capture and process them correctly.

Digital technologies have positively transformed society, but they have also led to undesirable consequences not anticipated at the time of design or development. We posit that insights into past undesirable consequences can help researchers and practitioners gain awareness and anticipate potential adverse effects. To test this assumption, we introduce BLIP, a system that extracts real-world undesirable consequences of technology from online articles, summarizes and categorizes them, and presents them in an interactive, web-based interface. In two user studies with 15 researchers in various computer science disciplines, we found that BLIP substantially increased the number and diversity of undesirable consequences they could list in comparison to relying on prior knowledge or searching online. Moreover, BLIP helped them identify undesirable consequences relevant to their ongoing projects, made them aware of undesirable consequences they "had never considered," and inspired them to reflect on their own experiences with technology.

As Transformers have become state-of-the-art models for natural language processing (NLP) tasks, the need to understand and explain their predictions is increasingly apparent. Especially in unsupervised applications, such as information retrieval tasks, similarity models built on top of foundation model representations have been widely applied. However, their inner prediction mechanisms have mostly remained opaque. Recent advances in explainable AI have made it possible to mitigate these limitations by leveraging improved explanations for Transformers through layer-wise relevance propagation (LRP). Using BiLRP, an extension developed for computing second-order explanations in bilinear similarity models, we investigate which feature interactions drive similarity in NLP models. We validate the resulting explanations and demonstrate their utility in three corpus-level use cases, analyzing grammatical interactions, multilingual semantics, and biomedical text retrieval. Our findings contribute to a deeper understanding of different semantic similarity tasks and models, highlighting how novel explainable AI methods enable in-depth analyses and corpus-level insights.

This paper explores the possibilities of the current generation of Large Language Models for incorporating Machine Learning Operations (MLOps) functionalities into ML training code bases. We evaluate the performance of OpenAI (gpt-3.5-turbo) and WizardCoder (open-source, 15B parameters) models on the automated accomplishment of various MLOps functionalities in different settings. We perform a benchmarking study that assesses the ability of these models to: (1) adapt existing code samples (Inlining) with component-specific MLOps functionality such as MLflow and Weights & Biases for experiment tracking, Optuna for hyperparameter optimization etc., and (2) perform the task of Translation from one component of an MLOps functionality to another, e.g., translating existing GitPython library based version control code to Data Version Control library based. We also propose three different approaches that involve teaching LLMs to comprehend the API documentation of the components as a reference while accomplishing the Translation tasks. In our evaluations, the gpt-3.5-turbo model significantly outperforms WizardCoder by achieving impressive Pass@3 accuracy in model optimization (55% compared to 0% by WizardCoder), experiment tracking (100%, compared to 62.5% by WizardCoder), model registration (92% compared to 42% by WizardCoder) and hyperparameter optimization (83% compared to 58% by WizardCoder) on average, in their best possible settings, showcasing its superior code adaptability performance in complex MLOps tasks.

Linear transformers have emerged as a subquadratic-time alternative to softmax attention and have garnered significant interest due to their fixed-size recurrent state that lowers inference cost. However, their original formulation suffers from poor scaling and underperforms compute-matched transformers. Recent linear models such as RWKV and Mamba have attempted to address these shortcomings by proposing novel time-mixing and gating architectures, but pre-training large language models requires significant data and compute investments. Thus, the search for subquadratic architectures is limited by the availability of compute and quality pre-training datasets. As a cost-effective alternative to pre-training linear transformers, we propose Scalable UPtraining for Recurrent Attention (SUPRA). We present a method to uptrain existing large pre-trained transformers into Recurrent Neural Networks (RNNs) with a modest compute budget. This allows us to leverage the strong pre-training data and performance of existing transformer LLMs, while requiring 5% of the training cost. We find that our linearization technique leads to competitive performance on standard benchmarks, but we identify persistent in-context learning and long-context modeling shortfalls for even the largest linear models. Our code and models can be found at https://github.com/TRI-ML/linear_open_lm.

We develop a methodology for analyzing language model task performance at the individual example level based on training data density estimation. Experiments with paraphrasing as a controlled intervention on finetuning data demonstrate that increasing the support in the training distribution for specific test queries results in a measurable increase in density, which is also a significant predictor of the performance increase caused by the intervention. Experiments with pretraining data demonstrate that we can explain a significant fraction of the variance in model perplexity via density measurements. We conclude that our framework can provide statistical evidence of the dependence of a target model's predictions on subsets of its training data, and can more generally be used to characterize the support (or lack thereof) in the training data for a given test task.