arXiv subjects

Chenyang Zhao

Publications and source records attributed to Chenyang Zhao.

67 records · Page 4Linked to original sources

arXiv2023

YAYI 2: Multilingual Open-Source Large Language Models

As the latest advancements in natural language processing, large language models (LLMs) have achieved human-level language understanding and generation abilities in many real-world tasks, and even have been regarded as a potential path to the artificial general intelligence. To better facilitate research on LLMs, many open-source LLMs, such as Llama 2 and Falcon, have recently been proposed and gained comparable performances to proprietary models. However, these models are primarily designed for English scenarios and exhibit poor performances in Chinese contexts. In this technical report, we propose YAYI 2, including both base and chat models, with 30 billion parameters. YAYI 2 is pre-trained from scratch on a multilingual corpus which contains 2.65 trillion tokens filtered by our pre-training data processing pipeline. The base model is aligned with human values through supervised fine-tuning with millions of instructions and reinforcement learning from human feedback. Extensive experiments on multiple benchmarks, such as MMLU and CMMLU, consistently demonstrate that the proposed YAYI 2 outperforms other similar sized open-source models.

cs.CL↗

arXiv2023

Prompt2Model: Generating Deployable Models from Natural Language Instructions

Large language models (LLMs) enable system builders today to create competent NLP systems through prompting, where they only need to describe the task in natural language and provide a few examples. However, in other ways, LLMs are a step backward from traditional special-purpose NLP models; they require extensive computational resources for deployment and can be gated behind APIs. In this paper, we propose Prompt2Model, a general-purpose method that takes a natural language task description like the prompts provided to LLMs, and uses it to train a special-purpose model that is conducive to deployment. This is done through a multi-step process of retrieval of existing datasets and pretrained models, dataset generation using LLMs, and supervised fine-tuning on these retrieved and generated datasets. Over three tasks, we demonstrate that given the same few-shot prompt as input, Prompt2Model trains models that outperform the results of a strong LLM, gpt-3.5-turbo, by an average of 20% while being up to 700 times smaller. We also show that this data can be used to obtain reliable performance estimates of model performance, enabling model developers to assess model reliability before deployment. Prompt2Model is available open-source at https://github.com/neulab/prompt2model.

cs.CL↗

arXiv2023

On Context Distribution Shift in Task Representation Learning for Offline Meta RL

Offline Meta Reinforcement Learning (OMRL) aims to learn transferable knowledge from offline datasets to enhance the learning process for new target tasks. Context-based Reinforcement Learning (RL) adopts a context encoder to expediently adapt the agent to new tasks by inferring the task representation, and then adjusting the policy based on this inferred representation. In this work, we focus on context-based OMRL, specifically on the challenge of learning task representation for OMRL. We conduct experiments that demonstrate that the context encoder trained on offline datasets might encounter distribution shift between the contexts used for training and testing. To overcome this problem, we present a hard-sampling-based strategy to train a robust task context encoder. Our experimental findings on diverse continuous control tasks reveal that utilizing our approach yields more robust task representations and better testing performance in terms of accumulated returns compared to baseline methods. Our code is available at https://github.com/ZJLAB-AMMI/HS-OMRL.

cs.LG↗

arXiv2023

Replicating Complex Dialogue Policy of Humans via Offline Imitation Learning with Supervised Regularization

Policy learning (PL) is a module of a task-oriented dialogue system that trains an agent to make actions in each dialogue turn. Imitating human action is a fundamental problem of PL. However, both supervised learning (SL) and reinforcement learning (RL) frameworks cannot imitate humans well. Training RL models require online interactions with user simulators, while simulating complex human policy is hard. Performances of SL-based models are restricted because of the covariate shift problem. Specifically, a dialogue is a sequential decision-making process where slight differences in current utterances and actions will cause significant differences in subsequent utterances. Therefore, the generalize ability of SL models is restricted because statistical characteristics of training and testing dialogue data gradually become different. This study proposed an offline imitation learning model that learns policy from real dialogue datasets and does not require user simulators. It also utilizes state transition information, which alleviates the influence of the covariate shift problem. We introduced a regularization trick to make our model can be effectively optimized. We investigated the performance of our model on four independent public dialogue datasets. The experimental result showed that our model performed better in the action prediction task.

cs.CL↗

arXiv2023

ODAM: Gradient-based instance-specific visual explanations for object detection

We propose the gradient-weighted Object Detector Activation Maps (ODAM), a visualized explanation technique for interpreting the predictions of object detectors. Utilizing the gradients of detector targets flowing into the intermediate feature maps, ODAM produces heat maps that show the influence of regions on the detector's decision for each predicted attribute. Compared to previous works classification activation maps (CAM), ODAM generates instance-specific explanations rather than class-specific ones. We show that ODAM is applicable to both one-stage detectors and two-stage detectors with different types of detector backbones and heads, and produces higher-quality visual explanations than the state-of-the-art both effectively and efficiently. We next propose a training scheme, Odam-Train, to improve the explanation ability on object discrimination of the detector through encouraging consistency between explanations for detections on the same object, and distinct explanations for detections on different objects. Based on the heat maps produced by ODAM with Odam-Train, we propose Odam-NMS, which considers the information of the model's explanation for each prediction to distinguish the duplicate detected objects. We present a detailed analysis of the visualized explanations of detectors and carry out extensive experiments to validate the effectiveness of the proposed ODAM.

cs.CV↗

arXiv2022

Towards Trustworthy Multi-label Sewer Defect Classification via Evidential Deep Learning

An automatic vision-based sewer inspection plays a key role of sewage system in a modern city. Recent advances focus on utilizing deep learning model to realize the sewer inspection system, benefiting from the capability of data-driven feature representation. However, the inherent uncertainty of sewer defects is ignored, resulting in the missed detection of serious unknown sewer defect categories. In this paper, we propose a trustworthy multi-label sewer defect classification (TMSDC) method, which can quantify the uncertainty of sewer defect prediction via evidential deep learning. Meanwhile, a novel expert base rate assignment (EBRA) is proposed to introduce the expert knowledge for describing reliable evidences in practical situations. Experimental results demonstrate the effectiveness of TMSDC and the superior capability of uncertainty estimation is achieved on the latest public benchmark.

cs.CV↗

arXiv2020

Robust Domain Randomised Reinforcement Learning through Peer-to-Peer Distillation

In reinforcement learning, domain randomisation is an increasingly popular technique for learning more general policies that are robust to domain-shifts at deployment. However, naively aggregating information from randomised domains may lead to high variance in gradient estimation and unstable learning process. To address this issue, we present a peer-to-peer online distillation strategy for RL termed P2PDRL, where multiple workers are each assigned to a different environment, and exchange knowledge through mutual regularisation based on Kullback-Leibler divergence. Our experiments on continuous control tasks show that P2PDRL enables robust learning across a wider randomisation distribution than baselines, and more robust generalisation to new environments at testing.

cs.LG↗

arXiv2019

Investigating Generalisation in Continuous Deep Reinforcement Learning

Deep Reinforcement Learning has shown great success in a variety of control tasks. However, it is unclear how close we are to the vision of putting Deep RL into practice to solve real world problems. In particular, common practice in the field is to train policies on largely deterministic simulators and to evaluate algorithms through training performance alone, without a train/test distinction to ensure models generalise and are not overfitted. Moreover, it is not standard practice to check for generalisation under domain shift, although robustness to such system change between training and testing would be necessary for real-world Deep RL control, for example, in robotics. In this paper we study these issues by first characterising the sources of uncertainty that provide generalisation challenges in Deep RL. We then provide a new benchmark and thorough empirical evaluation of generalisation challenges for state of the art Deep RL methods. In particular, we show that, if generalisation is the goal, then common practice of evaluating algorithms based on their training performance leads to the wrong conclusions about algorithm choice. Finally, we evaluate several techniques for improving generalisation and draw conclusions about the most robust techniques to date.

cs.LG↗

arXiv2017

Pressure and strain effects on the optical properties of K4 phosphorus

An investigation of the mechanical, electronic, and optical properties of the recently reported material K4 phosphorus was made in this work. The K4 phosphorus has been proved to be mechanically and dynamically stable up to 7 GPa under hydrostatic pressure. We compared the elastic anisotropy, average acoustic velocity and Debye temperature of K4 phosphorus at 0 and 7 GPa. The ideal tensile at large strains of K4 phosphorus was also examined, with the results showing that it would cleave under the tensile strength of 8.5 GPa with the strain of 0.3. In addition, the effect of tensile strain and pressure on optical properties and band gap were studied.

cond-mat.mtrl-sci↗

arXiv2017

Cylindrical vector resonant modes achieved in planar photonic crystal cavities with enlarged air-holes

We reveal a triangular-lattice planar photonic crystal supports Bloch modes with radially and azimuthally symmetric electric field distributions at the top band-edge of the first photonic band. Bifurcated from the corresponding Bloch modes, two cylindrical vector resonant modes are achieved by simply enlarging the central air-hole of the planar photonic crystal, which have high quality factors around 3,000 and small mode volume. The far-field radiations of the two resonant modes present high-quality cylindrical vector beam profiles. The resonant modes could be optimized by modifying the six nearest neighboring air-holes around the central defect. The cylindrically symmetric characteristics of the resonant mode's near- and far-fields might provide a new view to investigate light-matter interactions and device developments in planar photonic crystal cavities.

physics.optics↗

arXiv2017

Microwatts continuous-wave pumped second harmonic generation in few- and mono-layer GaSe

We demonstrate the first achievement of continuous-wave (CW) pumped second harmonic generation (SHG) in few- and mono-layer gallium selenide (GaSe) flakes, which are coated on silicon photonic crystal (PC) cavities. Because of ultrahigh second order nonlinearity of the two-dimensional (2D) GaSe and localized resonant mode in the PC cavity, SHG's pump power is greatly reduced to microwatts. In a nine-layer GaSe coated PC cavity, while the optical power inside the GaSe flake is only 1.5 percent of that in the silicon PC slab, the SHG in GaSe is more than 650 times stronger than the third harmonic generation in silicon slab, indicating 2D GaSe's great potentials to strengthen nonlinear processes in silicon photonics. Our study opens up a new view to expand 2D materials' optoelectronic applications in nonlinear regime and chip-integrated active devices.

physics.optics↗

arXiv2017

High performance graphene oxide-based humidity sensor integrated on a photonic crystal cavity

We report a high performance relative humidity (RH) microsensor by coating a few-layer graphene oxide (GO) flake over a photonic crystal (PC) cavity. Since the GO layer has a high water-activity and interacts with the evanescent cavity mode strongly, the exposure of GO-PC cavity in varied humidity levels results in significant resonant wavelength shifts, showing a slope of 0.68 nm/%RH in the range of 60%-85% RH. By interrogating the power variation of the cavity reflection, the microsensor presents an ultrahigh sensitivity exceeding 3.9 dB/%RH. Relying on the unimpeded permeation of water molecules through the GO interlayers and microscale distribution of the cavity mode, the integrated sensor has a response time less than 100 ms, which promises successful measurements of human breathing. Combining with the ease of fabrication, this high performance RH sensor has potential applications requiring optical access, device compactness, and fast dynamic response.

physics.optics↗

arXiv2014

Generation of vector beams in planar photonic crystal cavities with multiple missing-hole defects

We propose a novel method to generate vector beams in planar photonic crystal cavities with multiple missing-hole defects. Simulating the resonant modes in the cavities, we observe that the optical fields in each defect have different phase and polarization state distributions, which promise the compositions of vector beams by the scattered light from the defects. The far-field radiation patterns of the cavity modes calculated via the Sommerfeld diffraction theory present vector beams possessing hollow intensity profiles and polarization singularities. In addition, the extraction efficiencies of the vector beams from the cavities could be improved by modifying the air-holes surrounding the defects. This planar photonic crystal cavity-based vector beam generator may provide useful insights for the on-chip controlling of vector beams in their propagations and interactions with matter.

physics.optics↗