Grounded dialogue models generate responses that are grounded on certain concepts. Limited by the distribution of grounded dialogue data, models trained on such data face the transferability challenges in terms of the data distribution and the type o
f grounded concepts. To address the challenges, we propose the grounded minimal editing framework, which minimally edits existing responses to be grounded on the given concept. Focusing on personas, we propose Grounded Minimal Editor (GME), which learns to edit by disentangling and recombining persona-related and persona-agnostic parts of the response. To evaluate persona-grounded minimal editing, we present the PersonaMinEdit dataset, and experimental results show that GME outperforms competitive baselines by a large margin. To evaluate the transferability, we experiment on the test set of BlendedSkillTalk and show that GME can edit dialogue models responses to largely improve their persona consistency while preserving the use of knowledge and empathy.
The potential diagnostic applications of magnet-actuated capsules have been greatly increased in recent years. For most of these potential applications, accurate position control of the capsule have been highly demanding. However, the friction betwee
n the robot and the environment as well as the drag force from the tether play a significant role during the motion control of the capsule. Moreover, these forces especially the friction force are typically hard to model beforehand. In this paper, we first designed a magnet-actuated tethered capsule robot, where the driving magnet is mounted on the end of a robotic arm. Then, we proposed a learning-based approach to model the friction force between the capsule and the environment, with the goal of increasing the control accuracy of the whole system. Finally, several real robot experiments are demonstrated to showcase the effectiveness of our proposed approach.
Motion planning is critical to realize the autonomous operation of mobile robots. As the complexity and stochasticity of robot application scenarios increase, the planning capability of the classical hierarchical motion planners is challenged. In rec
ent years, with the development of intelligent computation technology, the deep reinforcement learning (DRL) based motion planning algorithm has gradually become a research hotspot due to its advantageous features such as not relying on the map prior, model-free, and unified global and local planning paradigms. In this paper, we provide a systematic review of various motion planning methods. First, we summarize the representative and cutting-edge algorithms for each submodule of the classical motion planning architecture and analyze their performance limitations. Subsequently, we concentrate on reviewing RL-based motion planning approaches, including RL optimization motion planners, map-free end-to-end methods that integrate sensing and decision-making, and multi-robot cooperative planning methods. Last but not least, we analyze the urgent challenges faced by these mainstream RL-based motion planners in detail, review some state-of-the-art works for these issues, and propose suggestions for future research.
With the potential applications of capsule robots in medical endoscopy, accurate dynamic control of the capsule robot is becoming more and more important. In the scale of a capsule robot, the friction between capsule and the environment plays an esse
ntial role in the dynamic model, which is usually difficult to model beforehand. In the paper, a tethered capsule robot system driven by a robot manipulator is built, where a strong magnetic Halbach array is mounted on the robots end-effector to adjust the state of the capsule. To increase the control accuracy, the friction between capsule and the environment is learned with demonstrated trajectories. With the learned friction model, experimental results demonstrate an improvement of 5.6% in terms of tracking error.
We apply the conformal bootstrap technique to study the $U(1)$ Dirac spin liquid (i.e. $N_f=4$ QED$_3$) and the newly proposed $N=7$ Stiefel liquid (i.e. a conjectured 3d non-Lagrangian CFT without supersymmetry). For the $N_f=4$ QED$_3$, we focus on
the monopole operator and ($SU(4)$ adjoint) fermion bilinear operator. We bootstrap their single correlators as well as the mixed correlators between them. We first discuss the bootstrap kinks from single correlators. Some exponents of these bootstrap kinks are close to the expected values of QED$_3$, but we provide clear evidence that they should not be identified as the QED$_3$. We then provide rigorous numerical bounds for the Dirac spin liquid and the $N=7$ Stiefel liquid to be stable critical phases on the triangular and kagome lattice. For the triangular and kagome Dirac spin liquid, the rigorous lower bounds of the monopole operators scaling dimension are $1.046$ and $1.105$, respectively. These bounds are consistent with the latest Monte Carlo results.
In this paper, we propose to combine pretrained language models with the modular dialogue paradigm for open-domain dialogue modeling. Our method, semantic-enhanced finetuning, instantiates conversation understanding, planning, and response generation
as a language model finetuning task. At inference, we disentangle semantic and token variations by specifying sampling methods and constraints for each module separately. For training and evaluation, we present X-Weibo, a Chinese multi-turn open-domain dialogue dataset with automatic annotation for emotions, DAs, and topical words. Experiments show that semantic-enhanced finetuning outperforms strong baselines on non-semantic and semantic metrics, improves the human-evaluated relevance, coherence, and informativeness, and exhibits considerable controllability over semantic variables.
Autoregressive models have been widely used in unsupervised text style transfer. Despite their success, these models still suffer from the content preservation problem that they usually ignore part of the source sentence and generate some irrelevant
words with strong styles. In this paper, we propose a Non-Autoregressive generator for unsupervised text Style Transfer (NAST), which alleviates the problem from two aspects. First, we observe that most words in the transferred sentence can be aligned with related words in the source sentence, so we explicitly model word alignments to suppress irrelevant words. Second, existing models trained with the cycle loss align sentences in two stylistic text spaces, which lacks fine-grained control at the word level. The proposed non-autoregressive generator focuses on the connections between aligned words, which learns the word-level transfer between styles. For experiments, we integrate the proposed generator into two base models and evaluate them on two style transfer tasks. The results show that NAST can significantly improve the overall performance and provide explainable word alignments. Moreover, the non-autoregressive generator achieves over 10x speedups at inference. Our codes are available at https://github.com/thu-coai/NAST.
This paper considers an intelligent reflecting surface (IRS) assisted multi-input multi-output (MIMO) power splitting (PS) based simultaneous wireless information and power transfer (SWIPT) system with multiple PS receivers (PSRs). The objective is t
o maximize the achievable data rate of the system by jointly optimizing the PS ratios at the PSRs, the active transmit beamforming (ATB) at the access point (AP), and the passive reflective beamforming (PRB) at the IRS, while the constraints on maximum transmission power at the AP, the reflective phase shift of each element at the IRS, the individual minimum harvested energy requirement of each PSR, and the domain of PS ratio of each PSR are all satisfied. For this unsolved problem, however, since the optimization variables are intricately coupled and the constraints are conflicting, the formulated problem is non-convex, and cannot be addressed by employing exist approaches directly. To this end, we propose a joint optimization framework to solve this problem. Particularly, we reformulate it as an equivalent form by employing the Lagrangian dual transform and the fractional programming transform, and decompose the transformed problem into several sub-problems. Then, we propose an alternate optimization algorithm by capitalizing on the dual sub-gradient method, the successive convex approximation method, and the penalty-based majorization-minimization approach, to solve the sub-problems iteratively, and obtain the optimal solutions in nearly closed-forms. Numerical simulation results verify the effectiveness of the IRS in SWIPT system and indicate that the proposed algorithm offers a substantial performance gain.
We estimate the effects on the decay constants of charmonium and on heavy meson masses due to the charm quark in the sea. Our goal is to understand whether for these quantities $N_f=2+1$ lattice QCD simulations provide results that can be compared wi
th experiments or whether $N_f=2+1+1$ QCD including the charm quark in the sea needs to be simulated. We consider two theories, $N_f=0$ QCD and QCD with $N_f=2$ charm quarks in the sea. The charm sea effects (due to two charm quarks) are estimated comparing the results obtained in these two theories, after matching them and taking the continuum limit. The absence of light quarks allows us to simulate the $N_f=2$ theory at lattice spacings down to $0.023$ fm that are crucial for reliable continuum extrapolations. We find that sea charm quark effects are below $1%$ for the decay constants of charmonium. Our results show that decoupling of charm works well up to energies of about $500$ MeV. We also compute the derivatives of the decay constants and meson masses with respect to the charm mass. For these quantities we again do not see a significant dynamical charm quark effect, albeit with a lower precision. For mesons made of a charm quark and a heavy antiquark, whose mass is twice that of the charm quark, sea effects are only about $0.1%$ in the ratio of vector to pseudoscalar masses.
Excited state contamination remains one of the most challenging sources of systematic uncertainty to control in lattice QCD calculations of nucleon matrix elements and form factors. Most lattice QCD collaborations advocate for the use of high-statist
ics calculations at large time separations ($t_{rm sep}gtrsim1$ fm) to combat the signal-to-noise degradation. In this work we demonstrate that, for the nucleon axial charge, $g_A$, the alternative strategy of utilizing a large number of relatively low-statistics calculations at short to medium time separations ($0.2lesssim t_{rm sep}lesssim1$ fm), combined with a multi-state analysis, provides a more robust and economical method of quantifying and controlling the excited state systematic uncertainty, including correlated late-time fluctuations that may bias the ground state. We show that two classes of excited states largely cancel in the ratio of the three-point to two-point functions, leaving the third class, the transition matrix elements, as the dominant source of contamination. On an $m_piapprox310$ MeV ensemble, we observe the expected exponential suppression of excited state contamination in the Feynman-Hellmann correlation function relative to the standard three-point function; the excited states of the regular three-point function reduce to the 1% level for $t_{rm sep} >2$ fm while, for the Feynman-Hellmann correlation function, they are suppressed to 1% at $t_{rm sep}approx1$ fm. Independent analyses of the three-point and Feynman-Hellmann correlators yield consistent results for the ground state. However, a combined analysis allows for a more detailed and robust understanding of the excited state contamination, improving the demonstration that the ground state parameters are stable against variations in the excited state model, the number of excited states, and the truncation of early-time or late-time numerical data.
