arXiv:2403.02839v3 »Full PDF »Recently, there has been a growing trend of utilizing Large Language Model
(LLM) to evaluate the quality of other LLMs. Many studies have employed
proprietary close-sourced models, especially GPT-4, as the evaluator.
Alternatively, other works have fine-tuned judge models based on open-source
LLMs as the evaluator. While the fine-tuned judge models are claimed to achieve
comparable evaluation capability with GPT-4, in this work, we conduct an
empirical study of judge models. Our findings indicate that although the
fine-tuned judge models achieve high performance on in-domain test sets, even
surpassing GPT-4, they underperform GPT-4 across several dimensions, including
generalizability, fairness, aspect-specific evaluation, and scalability. We
also reveal that the fine-tuned judge model inherently operates as a
task-specific classifier, consequently imposing the limitations. Finally, we
introduce a integrated method, leveraging GPT-4 to compensate for the
limitations and improve the fine-tuned judges. Experiment results show our
method achieves accuracy on par with GPT-4 with only 50% of the API expense.Abstract
Boosting LLM Translation Skills without General Ability Loss via
Rationale Distillation
arXiv:2410.13944v1 »Full PDF »Large Language Models (LLMs) have achieved impressive results across numerous
NLP tasks but still encounter difficulties in machine translation. Traditional
methods to improve translation have typically involved fine-tuning LLMs using
parallel corpora. However, vanilla fine-tuning often leads to catastrophic
forgetting of the instruction-following capabilities and alignment with human
preferences, compromising their broad general abilities and introducing
potential security risks. These abilities, which are developed using
proprietary and unavailable training data, make existing continual instruction
tuning methods ineffective. To overcome this issue, we propose a novel approach
called RaDis (Rationale Distillation). RaDis harnesses the strong generative
capabilities of LLMs to create rationales for training data, which are then
"replayed" to prevent forgetting. These rationales encapsulate general
knowledge and safety principles, acting as self-distillation targets to
regulate the training process. By jointly training on both reference
translations and self-generated rationales, the model can learn new translation
skills while preserving its overall general abilities. Extensive experiments
demonstrate that our method enhances machine translation performance while
maintaining the broader capabilities of LLMs across other tasks. This work
presents a pathway for creating more versatile LLMs that excel in specialized
tasks without compromising generality and safety.Abstract
Auto-Arena: Automating LLM Evaluations with Agent Peer Battles and
Committee Discussions
arXiv:2405.20267v4 »Full PDF »As LLMs continuously evolve, there is an urgent need for a reliable
evaluation method that delivers trustworthy results promptly. Currently, static
benchmarks suffer from inflexibility and unreliability, leading users to prefer
human voting platforms like Chatbot Arena. However, human evaluations require
significant manual effort. To address this, we propose the Auto-Arena, an
innovative framework that automates the entire evaluation process using
LLM-powered agents. Firstly, an LLM examiner generates questions. Then, two LLM
candidates engage in a multi-round peer battle based on individual questions,
aiming at revealing their true performance differences. Finally, a committee of
LLM judges collaboratively discusses and decides the winner, reducing bias and
enhancing fairness. During the peer battles, we observe intriguing scenarios
where the LLM candidates display competitive behaviors and even learn from the
opponents. In our extensive experiments involving 15 recent LLMs, Auto-Arena
shows a 92.14% correlation with human preferences, surpassing all previous
expert-annotated benchmarks without any manual efforts. As a result, Auto-Arena
offers a promising alternative to current human evaluation platforms for
evaluating LLMs automatically.Abstract
arXiv:2407.21783v2 »Full PDF »Modern artificial intelligence (AI) systems are powered by foundation models.
This paper presents a new set of foundation models, called Llama 3. It is a
herd of language models that natively support multilinguality, coding,
reasoning, and tool usage. Our largest model is a dense Transformer with 405B
parameters and a context window of up to 128K tokens. This paper presents an
extensive empirical evaluation of Llama 3. We find that Llama 3 delivers
comparable quality to leading language models such as GPT-4 on a plethora of
tasks. We publicly release Llama 3, including pre-trained and post-trained
versions of the 405B parameter language model and our Llama Guard 3 model for
input and output safety. The paper also presents the results of experiments in
which we integrate image, video, and speech capabilities into Llama 3 via a
compositional approach. We observe this approach performs competitively with
the state-of-the-art on image, video, and speech recognition tasks. The
resulting models are not yet being broadly released as they are still under
development.Abstract
SeaLLMs 3: Open Foundation and Chat Multilingual Large Language Models
for Southeast Asian Languages
arXiv:2407.19672v1 »Full PDF »Large Language Models (LLMs) have shown remarkable abilities across various
tasks, yet their development has predominantly centered on high-resource
languages like English and Chinese, leaving low-resource languages underserved.
To address this disparity, we present SeaLLMs 3, the latest iteration of the
SeaLLMs model family, tailored for Southeast Asian languages. This region,
characterized by its rich linguistic diversity, has lacked adequate language
technology support. SeaLLMs 3 aims to bridge this gap by covering a
comprehensive range of languages spoken in this region, including English,
Chinese, Indonesian, Vietnamese, Thai, Tagalog, Malay, Burmese, Khmer, Lao,
Tamil, and Javanese. Leveraging efficient language enhancement techniques and a
specially constructed instruction tuning dataset, SeaLLMs 3 significantly
reduces training costs while maintaining high performance and versatility. Our
model excels in tasks such as world knowledge, mathematical reasoning,
translation, and instruction following, achieving state-of-the-art performance
among similarly sized models. Additionally, we prioritized safety and
reliability by addressing both general and culture-specific considerations and
incorporated mechanisms to reduce hallucinations. This work underscores the
importance of inclusive AI, showing that advanced LLM capabilities can benefit
underserved linguistic and cultural communities.Abstract
MoGU: A Framework for Enhancing Safety of Open-Sourced LLMs While
Preserving Their Usability
arXiv:2405.14488v1 »Full PDF »Large Language Models (LLMs) are increasingly deployed in various
applications. As their usage grows, concerns regarding their safety are rising,
especially in maintaining harmless responses when faced with malicious
instructions. Many defense strategies have been developed to enhance the safety
of LLMs. However, our research finds that existing defense strategies lead LLMs
to predominantly adopt a rejection-oriented stance, thereby diminishing the
usability of their responses to benign instructions. To solve this problem, we
introduce the MoGU framework, designed to enhance LLMs' safety while preserving
their usability. Our MoGU framework transforms the base LLM into two variants:
the usable LLM and the safe LLM, and further employs dynamic routing to balance
their contribution. When encountering malicious instructions, the router will
assign a higher weight to the safe LLM to ensure that responses are harmless.
Conversely, for benign instructions, the router prioritizes the usable LLM,
facilitating usable and helpful responses. On various open-sourced LLMs, we
compare multiple defense strategies to verify the superiority of our MoGU
framework. Besides, our analysis provides key insights into the effectiveness
of MoGU and verifies that our designed routing mechanism can effectively
balance the contribution of each variant by assigning weights. Our work
released the safer Llama2, Vicuna, Falcon, Dolphin, and Baichuan2.Abstract
Data Trading Combination Auction Mechanism based on the Exponential
Mechanism
arXiv:2405.07336v1 »Full PDF »With the widespread application of machine learning technology in recent
years, the demand for training data has increased significantly, leading to the
emergence of research areas such as data trading. The work in this field is
still in the developmental stage. Different buyers have varying degrees of
demand for various types of data, and auctions play a role in such scenarios
due to their authenticity and fairness. Recent related work has proposed
combination auction mechanisms for different domains. However, such mechanisms
have not addressed the privacy concerns of buyers. In this paper, we design a
\textit{Data Trading Combination Auction Mechanism based on the exponential
mechanism} (DCAE) to protect buyers' bidding privacy from being leaked. We
apply the exponential mechanism to select the final settlement price for the
auction and generate a probability distribution based on the relationship
between the price and the revenue. In the experimental aspect, we consider the
selection of different mechanisms under two scenarios, and the experimental
results show that this method can ensure high auction revenue and protect
buyers' privacy from being violated.Abstract
Experimental quantum adversarial learning with programmable
superconducting qubits
Quantum computing promises to enhance machine learning and artificial
intelligence. Different quantum algorithms have been proposed to improve a wide
spectrum of machine learning tasks. Yet, recent theoretical works show that,
similar to traditional classifiers based on deep classical neural networks,
quantum classifiers would suffer from the vulnerability problem: adding tiny
carefully-crafted perturbations to the legitimate original data samples would
facilitate incorrect predictions at a notably high confidence level. This will
pose serious problems for future quantum machine learning applications in
safety and security-critical scenarios. Here, we report the first experimental
demonstration of quantum adversarial learning with programmable superconducting
qubits. We train quantum classifiers, which are built upon variational quantum
circuits consisting of ten transmon qubits featuring average lifetimes of 150
μs, and average fidelities of simultaneous single- and two-qubit gates
above 99.94% and 99.4% respectively, with both real-life images (e.g., medical
magnetic resonance imaging scans) and quantum data. We demonstrate that these
well-trained classifiers (with testing accuracy up to 99%) can be practically
deceived by small adversarial perturbations, whereas an adversarial training
process would significantly enhance their robustness to such perturbations. Our
results reveal experimentally a crucial vulnerability aspect of quantum
learning systems under adversarial scenarios and demonstrate an effective
defense strategy against adversarial attacks, which provide a valuable guide
for quantum artificial intelligence applications with both near-term and future
quantum devices.Abstract
PSE-Match: A Viewpoint-free Place Recognition Method with Parallel
Semantic Embedding
12 pages, 12 figures, accepted by IEEE Transactions on Intelligent
Transportation Systems
Accurate localization on autonomous driving cars is essential for autonomy
and driving safety, especially for complex urban streets and search-and-rescue
subterranean environments where high-accurate GPS is not available. However
current odometry estimation may introduce the drifting problems in long-term
navigation without robust global localization. The main challenges involve
scene divergence under the interference of dynamic environments and effective
perception of observation and object layout variance from different viewpoints.
To tackle these challenges, we present PSE-Match, a viewpoint-free place
recognition method based on parallel semantic analysis of isolated semantic
attributes from 3D point-cloud models. Compared with the original point cloud,
the observed variance of semantic attributes is smaller. PSE-Match incorporates
a divergence place learning network to capture different semantic attributes
parallelly through the spherical harmonics domain. Using both existing
benchmark datasets and two in-field collected datasets, our experiments show
that the proposed method achieves above 70% average recall with top one
retrieval and above 95% average recall with top ten retrieval cases. And
PSE-Match has also demonstrated an obvious generalization ability with a
limited training dataset.Abstract
Depth Information Guided Crowd Counting for Complex Crowd Scenes
9 pages, 8 figures. The paper is under consideration at Pattern
Recognition Letters
It is important to monitor and analyze crowd events for the sake of city
safety. In an EDOF (extended depth of field) image with a crowded scene, the
distribution of people is highly imbalanced. People far away from the camera
look much smaller and often occlude each other heavily, while people close to
the camera look larger. In such a case, it is difficult to accurately estimate
the number of people by using one technique. In this paper, we propose a Depth
Information Guided Crowd Counting (DigCrowd) method to deal with crowded EDOF
scenes. DigCrowd first uses the depth information of an image to segment the
scene into a far-view region and a near-view region. Then Digcrowd maps the
far-view region to its crowd density map and uses a detection method to count
the people in the near-view region. In addition, we introduce a new crowd
dataset that contains 1000 images. Experimental results demonstrate the
effectiveness of our DigCrowd methodAbstract
16737 latest Fairness/Ethics + ML/AI papers
