Shangmin Guo

Mr. Shangmin Guo

University of Edinburgh, Edinburgh, City of Edinburgh, UK
orcid

Research Focus & Metrics

Academic Career Length

5Years (2022-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
16 October 2024 in IEEE Transactions on Knowledge and Data Engineering
Verified
Journal Article
Closed access
WHWei Huang
Wei Huang
YZYunxiao Zhang
Yunxiao Zhang
SGShangmin Guo
Shangmin Guo
YSYu-Ming Shang
Yu-Ming Shang
XFXiangling Fu
Xiangling Fu
Selecting key data subsets for model training is an effective way to improve training efficiency. Existing methods generally utilize a well-trained model to evaluate samples and select crucial subsets, ignoring the fact that the sample importance changes dynamically during model training, resulting in the selected subset only being critical in a specific training epoch rather than a changing training phase. To address this issue, we attempt to evaluate the significant changes in sample importance during dynamic training and propose a novel data selection method to improve model training efficiency. Specifically, the temporal changes in sample importance are considered from three perspectives: (i) loss, the difference between the predicted labels and the true labels of samples in the current training epoch; (ii) instability, the dispersion of sample importance in the recent training phase; and (iii) inconsistency, the comparison of the changing trend in the importance of an individual sample relative to the average importance of all samples in the recent training phase. Extensive experiments demonstrate that dynamic data selection can reduce computational costs and improve model training efficiency. Additionally, we find that the difficulty level of the training task influences the data selection strategy.

17 September 2024 in Nature
Verified
Journal Article
Closed access
KSKenny Smith
Kenny Smith
SKSimon Kirby
Simon Kirby
SGShangmin Guo
Shangmin Guo
TGThomas L. Griffiths
Thomas L. Griffiths

26 October 2022 in Entropy
Verified
Journal Article
Open access
PDF file
BGBalint Gyevnar
Balint Gyevnar
GDGautier Dagan
Gautier Dagan
CHColeman Haley
Coleman Haley
SGShangmin Guo
Shangmin Guo
FMFrancis Mollica
Francis Mollica
Language evolution is driven by pressures for simplicity and informativity; however, the timescale on which these pressures operate is debated. Over several generations, learners’ biases for simple and informative systems can guide language evolution. Over repeated instances of dyadic communication, the principle of least effort dictates that speakers should bias systems towards simplicity and listeners towards informativity, similarly guiding language evolution. At the same time, it has been argued that learners only provide a bias for simplicity and, thus, language users must provide a bias for informativity. To what extent do languages evolve during acquisition versus use? We address this question by formally defining and investigating the communicative efficiency of acquisition trajectories. We illustrate our approach using colour-naming systems, replicating a communicative efficiency model based on the information bottleneck problem, and an acquisition model based on self-organising maps. We find that to the extent that language is iconic, learning alone is sufficient to shape language evolution. Regarding colour-naming systems specifically, we find that incorporating learning biases into communicative efficiency accounts might explain how speakers and listeners trade off communicative effort.

3 January 2024 in arXiv
Verified
Preprint
Open access
PDF file
SGShangmin Guo
Shangmin Guo
HBHaoran Bu
Haoran Bu
HWHaochuan Wang
Haochuan Wang
YRYi Ren
Yi Ren
DSDianbo Sui
Dianbo Sui
SLSiting Lu
Siting Lu
Large language models (LLMs) have been extensively used as the backbones for general-purpose agents, and some economics literature suggest that LLMs are capable of playing various types of economics games. Following these works, to overcome the limitation of evaluating LLMs using static benchmarks, we propose to explore competitive games as an evaluation for LLMs to incorporate multi-players and dynamicise the environment. By varying the game history revealed to LLMs-based players, we find that most of LLMs are rational in that they play strategies that can increase their payoffs, but not as rational as indicated by Nash Equilibria (NEs). Moreover, when game history are available, certain types of LLMs, such as GPT-4, can converge faster to the NE strategies, which suggests higher rationality level in comparison to other models. In the meantime, certain types of LLMs can win more often when game history are available, and we argue that the winning rate reflects the reasoning ability with respect to the strategies of other players. Throughout all our experiments, we observe that the ability to strictly follow the game rules described by natural languages also vary among the LLMs we tested. In this work, we provide an economics arena for the LLMs research community as a dynamic simulation to test the above-mentioned abilities of LLMs, i.e. rationality, strategic reasoning ability, and instruction-following capability.

29 February 2024 in arXiv
Verified
Preprint
Open access
PDF file
SGShangmin Guo
Shangmin Guo
BZBiao Zhang
Biao Zhang
TLTianlin Liu
Tianlin Liu
TLTianqi Liu
Tianqi Liu
MKMisha Khalman
Misha Khalman
MBMathieu Blondel
Mathieu Blondel
Direct alignment from preferences (DAP) methods, such as DPO, have recently emerged as efficient alternatives to reinforcement learning from human feedback (RLHF), that do not require a separate reward model. However, the preference datasets used in DAP methods are usually collected ahead of training and never updated, thus the feedback is purely offline. Moreover, responses in these datasets are often sampled from a language model distinct from the one being aligned, and since the model evolves over training, the alignment phase is inevitably off-policy. In this study, we posit that online feedback is key and improves DAP methods. Our method, online AI feedback (OAIF), uses an LLM as annotator: on each training iteration, we sample two responses from the current model and prompt the LLM annotator to choose which one is preferred, thus providing online feedback. Despite its simplicity, we demonstrate via human evaluation in several tasks that OAIF outperforms both offline DAP and RLHF methods. We further show that the feedback leveraged in OAIF is easily controllable, via instruction prompts to the LLM annotator.

11 December 2023 in arXiv
Verified
Preprint
Open access
PDF file
SGSamuel Garcin
Samuel Garcin
JDJames Doran
James Doran
SGShangmin Guo
Shangmin Guo
CLChristopher G. Lucas
Christopher G. Lucas
SAStefano V. Albrecht
Stefano V. Albrecht
A key limitation preventing the wider adoption of autonomous agents trained via deep reinforcement learning (RL) is their limited ability to generalise to new environments, even when these share similar characteristics with environments encountered during training. In this work, we investigate how a non-uniform sampling strategy of individual environment instances, or levels, affects the zero-shot generalisation (ZSG) ability of RL agents, considering two failure modes: overfitting and over-generalisation. As a first step, we measure the mutual information (MI) between the agent's internal representation and the set of training levels, which we find to be well-correlated to instance overfitting. In contrast to uniform sampling, adaptive sampling strategies prioritising levels based on their value loss are more effective at maintaining lower MI, which provides a novel theoretical justification for this class of techniques. We then turn our attention to unsupervised environment design (UED) methods, which adaptively generate new training levels and minimise MI more effectively than methods sampling from a fixed set. However, we find UED methods significantly shift the training distribution, resulting in over-generalisation and worse ZSG performance over the distribution of interest. To prevent both instance overfitting and over-generalisation, we introduce self-supervised environment design (SSED). SSED generates levels using a variational autoencoder, effectively reducing MI while minimising the shift with the distribution of interest, and leads to statistically significant improvements in ZSG over fixed-set level sampling strategies and UED methods.