Judging LLM-as-a-judge with MT-Bench and Chatbot Arena

Evaluating large language model (LLM) based chat assistants is challenging due to their broad capabilities and the inadequacy of existing benchmarks in measuring human preferences. To address this, we explore using strong LLMs as judges to evaluate these models on more open-ended questions. We examine the usage and limitations of LLM-as-a-judge, such as position and verbosity biases and limited reasoning ability, and propose solutions to migrate some of them. We then verify the agreement between LLM judges and human preferences by introducing two benchmarks: MT-bench, a multi-turn question set; and Chatbot Arena, a crowdsourced battle platform. Our results reveal that strong LLM judges like GPT-4 can match both controlled and crowdsourced human preferences well, achieving over 80\% agreement, the same level of agreement between humans. Hence, LLM-as-a-judge is a scalable and explainable way to approximate human preferences, which are otherwise very expensive to obtain. Additionally, we show our benchmark and traditional benchmarks complement each other by evaluating several variants of LLaMA/Vicuna. We will publicly release 80 MT-bench questions, 3K expert votes, and 30K conversations with human preferences from Chatbot Arena.

ShaDDR: Real-Time Example-Based Geometry and Texture Generation via 3D Shape Detailization and Differentiable Rendering

We present ShaDDR, an example-based deep generative neural network which produces a high-resolution textured 3D shape through geometry detailization and conditional texture generation applied to an input coarse voxel shape. Trained on a small set of detailed and textured exemplar shapes, our method learns to detailize the geometry via multi-resolution voxel upsampling and generate textures on voxel surfaces via differentiable rendering against exemplar texture images from a few views. The generation is real-time, taking less than 1 second to produce a 3D model with voxel resolutions up to 512^3. The generated shape preserves the overall structure of the input coarse voxel model, while the style of the generated geometric details and textures can be manipulated through learned latent codes. In the experiments, we show that our method can generate higher-resolution shapes with plausible and improved geometric details and clean textures compared to prior works. Furthermore, we showcase the ability of our method to learn geometric details and textures from shapes reconstructed from real-world photos. In addition, we have developed an interactive modeling application to demonstrate the generalizability of our method to various user inputs and the controllability it offers, allowing users to interactively sculpt a coarse voxel shape to define the overall structure of the detailized 3D shape.

FDNeRF: Semantics-Driven Face Reconstruction, Prompt Editing and Relighting with Diffusion Models

The ability to create high-quality 3D faces from a single image has become increasingly important with wide applications in video conferencing, AR/VR, and advanced video editing in movie industries. In this paper, we propose Face Diffusion NeRF (FDNeRF), a new generative method to reconstruct high-quality Face NeRFs from single images, complete with semantic editing and relighting capabilities. FDNeRF utilizes high-resolution 3D GAN inversion and expertly trained 2D latent-diffusion model, allowing users to manipulate and construct Face NeRFs in zero-shot learning without the need for explicit 3D data. With carefully designed illumination and identity preserving loss, as well as multi-modal pre-training, FD-NeRF offers users unparalleled control over the editing process enabling them to create and edit face NeRFs using just single-view images, text prompts, and explicit target lighting. The advanced features of FDNeRF have been designed to produce more impressive results than existing 2D editing approaches that rely on 2D segmentation maps for editable attributes. Experiments show that our FDNeRF achieves exceptionally realistic results and unprecedented flexibility in editing compared with state-of-the-art 3D face reconstruction and editing methods. Our code will be available at https://github.com/BillyXYB/FDNeRF.

BRIGHT: Bi-level Feature Representation of Image Collections using Groups of Hash Tables

We present BRIGHT, a bi-level feature representation for an image collection, consisting of a per-image latent space on top of a multi-scale feature grid space. Our representation is learned by an autoencoder to encode images into continuous key codes, which are used to retrieve features from groups of multi-resolution hash tables. Our key codes and hash tables are trained together continuously with well-defined gradient flows, leading to high usage of the hash table entries and improved generative modeling compared to discrete Vector Quantization (VQ). Differently from existing continuous representations such as KL-regularized latent codes, our key codes are strictly bounded in scale and variance. Overall, feature encoding by BRIGHT is compact, efficient to train, and enables generative modeling over the image codes using state-of-the-art generators such as latent diffusion models(LDMs). Experimental results show that our method achieves comparable reconstruction results to VQ methods while having a smaller and more efficient decoder network. By applying LDM over our key code space, we achieve state-of-the-art performance on image synthesis on the LSUN-Church and human-face datasets.

MS-DETR: Natural Language Video Localization with Sampling Moment-Moment Interaction

Given a query, the task of Natural Language Video Localization (NLVL) is to localize a temporal moment in an untrimmed video that semantically matches the query. In this paper, we adopt a proposal-based solution that generates proposals (i.e., candidate moments) and then select the best matching proposal. On top of modeling the cross-modal interaction between candidate moments and the query, our proposed Moment Sampling DETR (MS-DETR) enables efficient moment-moment relation modeling. The core idea is to sample a subset of moments guided by the learnable templates with an adopted DETR (DEtection TRansformer) framework. To achieve this, we design a multi-scale visual-linguistic encoder, and an anchor-guided moment decoder paired with a set of learnable templates. Experimental results on three public datasets demonstrate the superior performance of MS-DETR.

Semantic Segmentation with Bidirectional Language Models Improves Long-form ASR

We propose a method of segmenting long-form speech by separating semantically complete sentences within the utterance. This prevents the ASR decoder from needlessly processing faraway context while also preventing it from missing relevant context within the current sentence. Semantically complete sentence boundaries are typically demarcated by punctuation in written text; but unfortunately, spoken real-world utterances rarely contain punctuation. We address this limitation by distilling punctuation knowledge from a bidirectional teacher language model (LM) trained on written, punctuated text. We compare our segmenter, which is distilled from the LM teacher, against a segmenter distilled from a acoustic-pause-based teacher used in other works, on a streaming ASR pipeline. The pipeline with our segmenter achieves a 3.2% relative WER gain along with a 60 ms median end-of-segment latency reduction on a YouTube captioning task.

Probing Ring Resonator Sensor Based on Vernier Effect

The Vernier effect has seen extensive application in optical structures, serving to augment the free spectral range (FSR). A substantial FSR is vital in a myriad of applications including multiplexers, enabling a broad, clear band comparable to the C-band to accommodate a maximum number of channels. Nevertheless, a large FSR often conflicts with bending loss, as it necessitates a smaller resonator radius, thus increase the insertion loss in the bending portion. To facilitate FSR expansion without amplifying bending loss, we employed cascaded and parallel racetrack resonators and ring resonators of varying radius that demonstrate the Vernier effect. In this study, we designed, fabricated, and tested multiple types of racetrack resonators to validate the Vernier effect and its FSR extension capabilities. Our investigations substantiate that the Vernier effect, based on cascaded and series-coupled micro-ring resonator (MRR) sensors, can efficiently mitigate intra-channel cross-talk at higher data rates. This is achieved by providing larger input-to-through suppression, thus paving the way for future applications.

Efficient Detection of LLM-generated Texts with a Bayesian Surrogate Model

The detection of machine-generated text, especially from large language models (LLMs), is crucial in preventing serious social problems resulting from their misuse. Some methods train dedicated detectors on specific datasets but fall short in generalizing to unseen test data, while other zero-shot ones often yield suboptimal performance. Although the recent DetectGPT has shown promising detection performance, it suffers from significant inefficiency issues, as detecting a single candidate requires scoring hundreds of its perturbations with the source LLM. This paper aims to bridge this gap. Technically, we propose to incorporate a Bayesian surrogate model, which allows us to select typical samples based on Bayesian uncertainty and interpolate scores from typical samples to other ones, to improve query efficiency. Our empirical results demonstrate that our method significantly outperforms existing approaches under a low query budget. Notably, our method achieves similar performance with up to 2 times fewer queries than DetectGPT and 3.7% higher AUROC at a query number of 5.

NoisywikiHow: A Benchmark for Learning with Real-world Noisy Labels in Natural Language Processing

Large-scale datasets in the real world inevitably involve label noise. Deep models can gradually overfit noisy labels and thus degrade model generalization. To mitigate the effects of label noise, learning with noisy labels (LNL) methods are designed to achieve better generalization performance. Due to the lack of suitable datasets, previous studies have frequently employed synthetic label noise to mimic real-world label noise. However, synthetic noise is not instance-dependent, making this approximation not always effective in practice. Recent research has proposed benchmarks for learning with real-world noisy labels. However, the noise sources within may be single or fuzzy, making benchmarks different from data with heterogeneous label noises in the real world. To tackle these issues, we contribute NoisywikiHow, the largest NLP benchmark built with minimal supervision. Specifically, inspired by human cognition, we explicitly construct multiple sources of label noise to imitate human errors throughout the annotation, replicating real-world noise, whose corruption is affected by both ground-truth labels and instances. Moreover, we provide a variety of noise levels to support controlled experiments on noisy data, enabling us to evaluate LNL methods systematically and comprehensively. After that, we conduct extensive multi-dimensional experiments on a broad range of LNL methods, obtaining new and intriguing findings.

Deep Learning for Joint Acoustic Echo and Acoustic Howling Suppression in Hybrid Meetings

Hybrid meetings have become increasingly necessary during the post-COVID period and also brought new challenges for solving audio-related problems. In particular, the interplay between acoustic echo and acoustic howling in a hybrid meeting makes the joint suppression of them difficult. This paper proposes a deep learning approach to tackle this problem by formulating a recurrent feedback suppression process as an instantaneous speech separation task using the teacher-forced training strategy. Specifically, a self-attentive recurrent neural network is utilized to extract the target speech from microphone recordings with accessible and learned reference signals, thus suppressing acoustic echo and acoustic howling simultaneously. Different combinations of input signals and loss functions have been investigated for performance improvement. Experimental results demonstrate the effectiveness of the proposed method for suppressing echo and howling jointly in hybrid meetings.