Unlock the Correlation between Supervised Fine-Tuning and Reinforcement Learning in Training Code Large Language Models

Automatic code generation has been a longstanding research topic. With the advancement of general-purpose large language models (LLMs), the ability to code stands out as one important measure to the model's reasoning performance. Usually, a two-stage training paradigm is implemented to obtain a Code LLM, namely the pretraining and the fine-tuning. Within the fine-tuning, supervised fine-tuning (SFT), and reinforcement learning (RL) are often used to improve the model's zero-shot ability. A large number of work has been conducted to improve the model's performance on code-related benchmarks with either modifications to the algorithm or refinement of the dataset. However, we still lack a deep insight into the correlation between SFT and RL. For instance, what kind of dataset should be used to ensure generalization, or what if we abandon the SFT phase in fine-tuning. In this work, we make an attempt to understand the correlation between SFT and RL. To facilitate our research, we manually craft 100 basis python functions, called atomic functions, and then a synthesizing pipeline is deployed to create a large number of synthetic functions on top of the atomic ones. In this manner, we ensure that the train and test sets remain distinct, preventing data contamination. Through comprehensive ablation study, we find: (1) Both atomic and synthetic functions are indispensable for SFT's generalization, and only a handful of synthetic functions are adequate; (2) Through RL, the SFT's generalization to target domain can be greatly enhanced, even with the same training prompts; (3) Training RL from scratch can alleviate the over-fitting issue introduced in the SFT phase.

Mistral-C2F: Coarse to Fine Actor for Analytical and Reasoning Enhancement in RLHF and Effective-Merged LLMs

Despite the advances in Large Language Models (LLMs), exemplified by models like GPT-4 and Claude, smaller-scale LLMs such as Llama and Mistral often struggle with generating in-depth and coherent dialogues. This paper presents a novel two-step Coarse-to-Fine Actor model to address the inherent limitations in conversational and analytical capabilities of small-sized LLMs. Our approach begins with the Policy-based Coarse Actor, employing a technique we term "Continuous Maximization". The Coarse Actor establishes an enhanced, knowledge-rich pool adept at aligning with human preference styles in analysis and reasoning. Through the RLHF process, it employs Continuous Maximization, a strategy that dynamically and adaptively extends the output length limit, enabling the generation of more detailed and analytical content. Subsequently, the Fine Actor refines this analytical content, addressing the generation of excessively redundant information from the Coarse Actor. We introduce a "Knowledge Residue Merger" approach, refining the content from the Coarse Actor and merging it with an existing Instruction model to improve quality, correctness, and reduce redundancies. We applied our methodology to the popular Mistral model, creating Mistral-C2F, which has demonstrated exceptional performance across 11 general language tasks and the MT-Bench Dialogue task, outperforming similar-scale models and even larger models with 13B and 30B parameters. Our model has significantly improved conversational and analytical reasoning abilities.

Seeing the Image: Prioritizing Visual Correlation by Contrastive Alignment

Existing image-text modality alignment in Vision Language Models (VLMs) treats each text token equally in an autoregressive manner. Despite being simple and effective, this method results in sub-optimal cross-modal alignment by over-emphasizing the text tokens that are less correlated with or even contradictory with the input images. In this paper, we advocate for assigning distinct contributions for each text token based on its visual correlation. Specifically, we present by contrasting image inputs, the difference in prediction logits on each text token provides strong guidance of visual correlation. We therefore introduce Contrastive ALignment (CAL), a simple yet effective re-weighting strategy that prioritizes training visually correlated tokens. Our experimental results demonstrate that CAL consistently improves different types of VLMs across different resolutions and model sizes on various benchmark datasets. Importantly, our method incurs minimal additional computational overhead, rendering it highly efficient compared to alternative data scaling strategies. Codes are available at https://github.com/foundation-multimodal-models/CAL.

Efficient Text-driven Motion Generation via Latent Consistency Training

Motion diffusion models have recently proven successful for text-driven human motion generation. Despite their excellent generation performance, they are challenging to infer in real time due to the multi-step sampling mechanism that involves tens or hundreds of repeat function evaluation iterations. To this end, we investigate a motion latent consistency Training (MLCT) for motion generation to alleviate the computation and time consumption during iteration inference. It applies diffusion pipelines to low-dimensional motion latent spaces to mitigate the computational burden of each function evaluation. Explaining the diffusion process with probabilistic flow ordinary differential equation (PF-ODE) theory, the MLCT allows extremely few steps infer between the prior distribution to the motion latent representation distribution via maintaining consistency of the outputs over the trajectory of PF-ODE. Especially, we introduce a quantization constraint to optimize motion latent representations that are bounded, regular, and well-reconstructed compared to traditional variational constraints. Furthermore, we propose a conditional PF-ODE trajectory simulation method, which improves the conditional generation performance with minimal additional training costs. Extensive experiments on two human motion generation benchmarks show that the proposed model achieves state-of-the-art performance with less than 10\% time cost.

NeRF2Points: Large-Scale Point Cloud Generation From Street Views' Radiance Field Optimization

Neural Radiance Fields (NeRF) have emerged as a paradigm-shifting methodology for the photorealistic rendering of objects and environments, enabling the synthesis of novel viewpoints with remarkable fidelity. This is accomplished through the strategic utilization of object-centric camera poses characterized by significant inter-frame overlap. This paper explores a compelling, alternative utility of NeRF: the derivation of point clouds from aggregated urban landscape imagery. The transmutation of street-view data into point clouds is fraught with complexities, attributable to a nexus of interdependent variables. First, high-quality point cloud generation hinges on precise camera poses, yet many datasets suffer from inaccuracies in pose metadata. Also, the standard approach of NeRF is ill-suited for the distinct characteristics of street-view data from autonomous vehicles in vast, open settings. Autonomous vehicle cameras often record with limited overlap, leading to blurring, artifacts, and compromised pavement representation in NeRF-based point clouds. In this paper, we present NeRF2Points, a tailored NeRF variant for urban point cloud synthesis, notable for its high-quality output from RGB inputs alone. Our paper is supported by a bespoke, high-resolution 20-kilometer urban street dataset, designed for point cloud generation and evaluation. NeRF2Points adeptly navigates the inherent challenges of NeRF-based point cloud synthesis through the implementation of the following strategic innovations: (1) Integration of Weighted Iterative Geometric Optimization (WIGO) and Structure from Motion (SfM) for enhanced camera pose accuracy, elevating street-view data precision. (2) Layered Perception and Integrated Modeling (LPiM) is designed for distinct radiance field modeling in urban environments, resulting in coherent point cloud representations.

Balancing Enhancement, Harmlessness, and General Capabilities: Enhancing Conversational LLMs with Direct RLHF

In recent advancements in Conversational Large Language Models (LLMs), a concerning trend has emerged, showing that many new base LLMs experience a knowledge reduction in their foundational capabilities following Supervised Fine-Tuning (SFT). This process often leads to issues such as forgetting or a decrease in the base model's abilities. Moreover, fine-tuned models struggle to align with user preferences, inadvertently increasing the generation of toxic outputs when specifically prompted. To overcome these challenges, we adopted an innovative approach by completely bypassing SFT and directly implementing Harmless Reinforcement Learning from Human Feedback (RLHF). Our method not only preserves the base model's general capabilities but also significantly enhances its conversational abilities, while notably reducing the generation of toxic outputs. Our approach holds significant implications for fields that demand a nuanced understanding and generation of responses, such as customer service. We applied this methodology to Mistral, the most popular base model, thereby creating Mistral-Plus. Our validation across 11 general tasks demonstrates that Mistral-Plus outperforms similarly sized open-source base models and their corresponding instruct versions. Importantly, the conversational abilities of Mistral-Plus were significantly improved, indicating a substantial advancement over traditional SFT models in both safety and user preference alignment.

Referee-Meta-Learning for Fast Adaptation of Locational Fairness

When dealing with data from distinct locations, machine learning algorithms tend to demonstrate an implicit preference of some locations over the others, which constitutes biases that sabotage the spatial fairness of the algorithm. This unfairness can easily introduce biases in subsequent decision-making given broad adoptions of learning-based solutions in practice. However, locational biases in AI are largely understudied. To mitigate biases over locations, we propose a locational meta-referee (Meta-Ref) to oversee the few-shot meta-training and meta-testing of a deep neural network. Meta-Ref dynamically adjusts the learning rates for training samples of given locations to advocate a fair performance across locations, through an explicit consideration of locational biases and the characteristics of input data. We present a three-phase training framework to learn both a meta-learning-based predictor and an integrated Meta-Ref that governs the fairness of the model. Once trained with a distribution of spatial tasks, Meta-Ref is applied to samples from new spatial tasks (i.e., regions outside the training area) to promote fairness during the fine-tune step. We carried out experiments with two case studies on crop monitoring and transportation safety, which show Meta-Ref can improve locational fairness while keeping the overall prediction quality at a similar level.

ICE-GRT: Instruction Context Enhancement by Generative Reinforcement based Transformers

The emergence of Large Language Models (LLMs) such as ChatGPT and LLaMA encounter limitations in domain-specific tasks, with these models often lacking depth and accuracy in specialized areas, and exhibiting a decrease in general capabilities when fine-tuned, particularly analysis ability in small sized models. To address these gaps, we introduce ICE-GRT, utilizing Reinforcement Learning from Human Feedback (RLHF) grounded in Proximal Policy Optimization (PPO), demonstrating remarkable ability in in-domain scenarios without compromising general task performance. Our exploration of ICE-GRT highlights its understanding and reasoning ability to not only generate robust answers but also to provide detailed analyses of the reasons behind the answer. This capability marks a significant progression beyond the scope of Supervised Fine-Tuning models. The success of ICE-GRT is dependent on several crucial factors, including Appropriate Data, Reward Size Scaling, KL-Control, Advantage Normalization, etc. The ICE-GRT model exhibits state-of-the-art performance in domain-specific tasks and across 12 general Language tasks against equivalent size and even larger size LLMs, highlighting the effectiveness of our approach. We provide a comprehensive analysis of the ICE-GRT, underscoring the significant advancements it brings to the field of LLM.

SpatialRank: Urban Event Ranking with NDCG Optimization on Spatiotemporal Data

The problem of urban event ranking aims at predicting the top-k most risky locations of future events such as traffic accidents and crimes. This problem is of fundamental importance to public safety and urban administration especially when limited resources are available. The problem is, however, challenging due to complex and dynamic spatio-temporal correlations between locations, uneven distribution of urban events in space, and the difficulty to correctly rank nearby locations with similar features. Prior works on event forecasting mostly aim at accurately predicting the actual risk score or counts of events for all the locations. Rankings obtained as such usually have low quality due to prediction errors. Learning-to-rank methods directly optimize measures such as Normalized Discounted Cumulative Gain (NDCG), but cannot handle the spatiotemporal autocorrelation existing among locations. In this paper, we bridge the gap by proposing a novel spatial event ranking approach named SpatialRank. SpatialRank features adaptive graph convolution layers that dynamically learn the spatiotemporal dependencies across locations from data. In addition, the model optimizes through surrogates a hybrid NDCG loss with a spatial component to better rank neighboring spatial locations. We design an importance-sampling with a spatial filtering algorithm to effectively evaluate the loss during training. Comprehensive experiments on three real-world datasets demonstrate that SpatialRank can effectively identify the top riskiest locations of crimes and traffic accidents and outperform state-of-art methods in terms of NDCG by up to 12.7%.

GPT-Fathom: Benchmarking Large Language Models to Decipher the Evolutionary Path towards GPT-4 and Beyond

With the rapid advancement of large language models (LLMs), there is a pressing need for a comprehensive evaluation suite to assess their capabilities and limitations. Existing LLM leaderboards often reference scores reported in other papers without consistent settings and prompts, which may inadvertently encourage cherry-picking favored settings and prompts for better results. In this work, we introduce GPT-Fathom, an open-source and reproducible LLM evaluation suite built on top of OpenAI Evals. We systematically evaluate 10+ leading LLMs as well as OpenAI's legacy models on 20+ curated benchmarks across 7 capability categories, all under aligned settings. Our retrospective study on OpenAI's earlier models offers valuable insights into the evolutionary path from GPT-3 to GPT-4. Currently, the community is eager to know how GPT-3 progressively improves to GPT-4, including technical details like whether adding code data improves LLM's reasoning capability, which aspects of LLM capability can be improved by SFT and RLHF, how much is the alignment tax, etc. Our analysis sheds light on many of these questions, aiming to improve the transparency of advanced LLMs.