Beyond Model Collapse: Scaling Up with Synthesized Data Requires Reinforcement

Synthesized data from generative models is increasingly considered as an alternative to human-annotated data for fine-tuning Large Language Models. This raises concerns about model collapse: a drop in performance of models fine-tuned on generated data. Considering that it is easier for both humans and machines to tell between good and bad examples than to generate high-quality samples, we investigate the use of feedback on synthesized data to prevent model collapse. We derive theoretical conditions under which a Gaussian mixture classification model can achieve asymptotically optimal performance when trained on feedback-augmented synthesized data, and provide supporting simulations for finite regimes. We illustrate our theoretical predictions on two practical problems: computing matrix eigenvalues with transformers and news summarization with large language models, which both undergo model collapse when trained on model-generated data. We show that training from feedback-augmented synthesized data, either by pruning incorrect predictions or by selecting the best of several guesses, can prevent model collapse, validating popular approaches like RLHF.

A Tale of Tails: Model Collapse as a Change of Scaling Laws

As AI model size grows, neural scaling laws have become a crucial tool to predict the improvements of large models when increasing capacity and the size of original (human or natural) training data. Yet, the widespread use of popular models means that the ecosystem of online data and text will co-evolve to progressively contain increased amounts of synthesized data. In this paper we ask: How will the scaling laws change in the inevitable regime where synthetic data makes its way into the training corpus? Will future models, still improve, or be doomed to degenerate up to total (model) collapse? We develop a theoretical framework of model collapse through the lens of scaling laws. We discover a wide range of decay phenomena, analyzing loss of scaling, shifted scaling with number of generations, the ''un-learning" of skills, and grokking when mixing human and synthesized data. Our theory is validated by large-scale experiments with a transformer on an arithmetic task and text generation using the large language model Llama2.

A Joint Design for Full-duplex OFDM AF Relay System with Precoded Short Guard Interval

In-band full-duplex relay (FDR) has attracted much attention as an effective solution to improve the coverage and spectral efficiency in wireless communication networks. The basic problem for FDR transmission is how to eliminate the inherent self-interference and re-use the residual self-interference (RSI) at the relay to improve the end-to-end performance. Considering the RSI at the FDR, the overall equivalent channel can be modeled as an infinite impulse response (IIR) channel. For this IIR channel, a joint design for precoding, power gain control and equalization of cooperative OFDM relay systems is presented. Compared with the traditional OFDM systems, the length of the guard interval for the proposed design can be distinctly reduced, thereby improving the spectral efficiency. By analyzing the noise sources, this paper evaluates the signal to noise ratio (SNR) of the proposed scheme and presents a power gain control algorithm at the FDR. Compared with the existing schemes, the proposed scheme shows a superior bit error rate (BER) performance.

L2SR: Learning to Sample and Reconstruct for Accelerated MRI

Accelerated MRI aims to find a pair of samplers and reconstructors to reduce acquisition time while maintaining the reconstruction quality. Most of the existing works focus on finding either sparse samplers with a fixed reconstructor or finding reconstructors with a fixed sampler. Recently, people have begun to consider learning samplers and reconstructors jointly. In this paper, we propose an alternating training framework for finding a good pair of samplers and reconstructors via deep reinforcement learning (RL). In particular, we propose a novel sparse-reward Partially Observed Markov Decision Process (POMDP) to formulate the MRI sampling trajectory. Compared to the existing works that utilize dense-reward POMDPs, the proposed sparse-reward POMDP is more computationally efficient and has a provable advantage over dense-reward POMDPs. We evaluate our method on fastMRI, a public benchmark MRI dataset, and it achieves state-of-the-art reconstruction performances.

Deep Multi-Modal Sets

Many vision-related tasks benefit from reasoning over multiple modalities to leverage complementary views of data in an attempt to learn robust embedding spaces. Most deep learning-based methods rely on a late fusion technique whereby multiple feature types are encoded and concatenated and then a multi layer perceptron (MLP) combines the fused embedding to make predictions. This has several limitations, such as an unnatural enforcement that all features be present at all times as well as constraining only a constant number of occurrences of a feature modality at any given time. Furthermore, as more modalities are added, the concatenated embedding grows. To mitigate this, we propose Deep Multi-Modal Sets: a technique that represents a collection of features as an unordered set rather than one long ever-growing fixed-size vector. The set is constructed so that we have invariance both to permutations of the feature modalities as well as to the cardinality of the set. We will also show that with particular choices in our model architecture, we can yield interpretable feature performance such that during inference time we can observe which modalities are most contributing to the prediction.With this in mind, we demonstrate a scalable, multi-modal framework that reasons over different modalities to learn various types of tasks. We demonstrate new state-of-the-art performance on two multi-modal datasets (Ads-Parallelity [34] and MM-IMDb [1]).