Xiaochen Gao, Feng Yao, Kewen Zhao, Beilei He, Animesh Kumar, Vish Krishnan, Jingbo Shang. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

In order to enhance 5G capability, 3GPP proposed the expansion of NR operation into the unlicensed spectrum (NR-U). However, channel access in the unlicensed band is challenging due to other coexisting technologies such as WiFi and LTE Licence Assisted Access (LAA). Unlicensed channel access protocols like Listen Before Talk (LBT) can degrade the NR performance compared to the licensed case. Prior work has shown that using Bandwidth Parts (BWP) in NR-U can help alleviate the situation. The concept of BWP was introduced in 5G in order to flexibly provide differentiated services based on varying requirements. However, dynamically deciding to which BWP a UE should be assigned still remains a challenge. In this work, we propose two solutions to effectively assign BWPs in order to decrease the Head of Line (HoL) delay and increase the throughput for NR-U. The first approach, Least Collision Assignment (LCA), uses a provably optimal heuristic under certain conditions. The second one uses a Deep Recurrent Q-Network (DRQN) based reinforcement learning model for assignment (RLA). Our results show significant improvements in delay and throughput of 60.27% and 70.88% with LCA, and 36.37% and 19.42% with RLA.

The opacity in developing large language models (LLMs) is raising growing concerns about the potential contamination of public benchmarks in the pre-training data.Existing contamination detection methods are typically based on the text overlap between training and evaluation data, which can be too superficial to reflect deeper forms of contamination.In this paper, we first present a cross-lingual form of contamination that inflates LLMs' performance while evading current detection methods, deliberately injected by overfitting LLMs on the translated versions of benchmark test sets.Then, we propose generalization-based approaches to unmask such deeply concealed contamination.Specifically, we examine the LLM's performance change after modifying the original benchmark by replacing the false answer choices with correct ones from other questions.Contaminated models can hardly generalize to such easier situations, where the false choices can be not even wrong, as all choices are correct in their memorization.Experimental results demonstrate that cross-lingual contamination can easily fool existing detection methods, but not ours.In addition, we discuss the potential utilization of cross-lingual contamination in interpreting LLMs' working mechanisms and in post-training LLMs for enhanced multilingual capabilities.The code and dataset we use can be obtained from https:// github.com/ShangDataLab/Deep-Contam. * Equal contribution. Listing order is random.Para Sócrates, el alma se daña por la falta de __A__.A. conocimiento, B. riqueza C. comunidad, D. coraje For Socrates, the soul is harmed by lack of _A_. A. knowledge, B. wealth C.

Several distributed real-time signal sensing/monitoring systems require quantization for efficient signal representation. These distributed sensors often have computational and energy limitations. Motivated by this concern, we propose a novel quantization scheme called Approximate Lloyd-Max (ALM) that is nearly-optimal. Assuming a continuous and finite support probability distribution of the source, we show that our ALM quantizer converges to the classical Lloyd-Max quantizer with increasing bitrate. Our ALM quantizer, which is recursive, converges exponentially fast with the number of iteration. We illustrate our results using simulations for the Beta(4,2) distribution on the source.

Measurement of spatial fields is of interest in environment monitoring. Recently mobile sensing has been proposed for spatial field reconstruction, which requires a smaller number of sensors when compared to the traditional paradigm of sensing with static sensors. A challenge in mobile sensing is to overcome the location uncertainty of its sensors. While GPS or other localization methods can reduce this uncertainty, we address a more fundamental question: can a location-unaware mobile sensor, recording samples on a directed non-uniform random walk, learn the statistical distribution (as a function of space) of an underlying random process (spatial field)? The answer is in the affirmative for Lipschitz continuous fields, where the accuracy of our distribution-learning method increases with the number of observed field samples (sampling rate). To validate our distribution-learning method, we have created a dataset with 43 experimental trials by measuring sound-level along a fixed path using a location-unaware mobile sound-level meter.

Sampling of smooth spatiotemporally varying fields is a well-studied topic in the literature. Classical approach assumes that the field is observed at known sampling locations and known timestamps ensuring field reconstruction. In a first, in this work the sampling and reconstruction of a spatiotemporal bandlimited field is addressed, where the samples are obtained by a location-unaware, time-unaware mobile sensor. The spatial and temporal order of samples is assumed to be known. It is assumed that the field samples are affected by measurement-noise. The spatial field's evolution is modeled by a linear constant coefficient partial differential equation. A regression style estimate is developed for reconstruction of the spatial field. The intersample spacings and the intersample timestamp differences are assumed to be from independent unknown renewal processes. If n is the average number of samples of the field obtained by the mobile sensor, then it is shown that the mean-squared error decreases as O(1/n).

Sampling of physical fields with mobile sensor is an emerging area. In this context, this work introduces and addresses some aspects of a fundamental question: can a spatial field be estimated from samples taken at unknown sampling locations? In a field (signal) sampling setup, unknown sampling locations, sample quantization, unknown bandwidth of the field, and presence of measurement-noise present difficulties in the process of field estimation. In this work, except for quantization, the other three issues will be tackled together in a mobile-sampling framework. Spatially bandlimited fields are considered. It is assumed that measurement-noise affected field samples are collected on spatial locations obtained from an unknown renewal process. That is, the samples are obtained on locations obtained from a renewal process, but the sampling locations and the renewal process distribution are unknown. In this challenging unknown sampling location setup, it is shown that the mean-squared error in field estimation decreases as O(1/n) where n is the average number of samples collected by the mobile sensor. The average number of samples collected is determined by the inter-sample spacing distribution in the renewal process. An algorithm to ascertain spatial field's bandwidth is detailed, which works with high probability as the average number of samples n increases. This algorithm works in the same setup, i.e., in the presence of measurement-noise and unknown sampling locations.

Sampling spatial fields using sensors which are location unaware is an exciting topic. Due to symmetry and shift-invariance of bandlimited fields, it is known that uniformly distributed location-unaware sensors cannot infer the field. This work studies asymmetric (nonuniform) distributions on location-unaware sensors that will enable bandlimited field inference. In this first exposition, to facilitate analysis, location-unaware sensors are restricted to a discrete grid. Oversampling is used to overcome the lack of location information. The samples obtained from location-unaware sensors are clustered together to infer the field using the probability distribution that governs sensor placement on the grid. Based on this clustering algorithm, the main result of this work is to find the optimal probability distribution on sensor locations that minimizes the detection error-probability of the underlying spatial field.

Spatial field sampling with mobile sensor has recently been addressed in the literature. This work introduces and proposes solution to a fundamental question: can a spatial field be estimated from samples taken by a mobile sensor at unknown sampling locations along a path? Spatially one-dimensional and bandlimited, and temporally fixed fields are considered. It is assumed that field samples are collected on spatial locations realized by an unknown renewal process. That is, the sampling locations and the inter-sample distribution in the renewal process are both unknown. It is shown that average mean-squared error in field estimation decreases as O(1/n) where n is the sampling-rate employed by the mobile sensor. The sampling rate can be increased by controlling the mean value of the inter-sample spacing by a location unaware mobile sensor.

The problem of identifying the channel code from a received sequence of noise-affected codewords is known as the blind reconstruction of channel codes. Blind reconstruction of channel codes is an important problem in military surveillance applications to identify the channel code used by an adversary. In this paper, we consider the problem of the blind reconstruction of the binary cyclic codes of unknown length from an unsynchronized bitstream (i.e., when the location of codeword boundaries is not known). For the blind reconstruction of cyclic codes, it is sufficient to identify the correct synchronization, the length, and the factors of the generator polynomial of the code. Toward this, we study the distribution of the syndromes (remainders) of the received polynomials with respect to a candidate factor of the generator polynomial. We prove that the probability of zero syndrome is maximum when all the parameters are correct. Using this result, the problem of the blind reconstruction of cyclic codes is formulated and solved as a hypothesis testing problem.

Acquisition of physical fields, such as temperature along a path, using a distributed array of sensors (samples) is of interest. For smooth spatial fields, in a Nyquist style sampling setup, the aliasing error is determined by the (spatial) spectral profile of a field. Physical fields and their spectral properties evolve with time. In this work, the spectral evolution of spatio-temporal fields is analyzed, where the field is given by physical law comprising of a constant coefficient linear partial differential equation. A procedure to examine whether field's spatial spectral profile will become worse with time, from aliasing point of view, is developed. The procedure is exemplified using a second-order PDE in this work. The analysis is extended to include simple point source terms. Techniques such as the Fourier transform, the unilateral Laplace transform, and root-locus plot from control theory are utilized in this work.

Unlike in the developed countries, TV white spaces are abundant in India and rural broadband coverage is negligible. Government of India has an ambitious plan to connect 250,000 rural offices (Gram Panchayat) using optical fiber based point of presence (PoP); this will allay, but not solve the rural broadband coverage problem. In this work, a middle-mile multihop mesh network operating in the TV UHF band is presented as a solution to providing seamless connectivity between the Gram Panchayat and village users. This network can coexist with TV broadcasting via the License Shared Access mechanism. The main contribution of this paper is a throughput optimization tool. Given the demography information of rural areas, the desired broadband speed and contention ratio, available frequency, location of fiber PoP, and propagation path-loss models of the TV UHF band - our optimization tool computes optimal power and routing for the multihop mesh network. The analysis is performed at the physical layer. Time division multiple access and spatial reuse of frequency are used for interference management in this work. Simply put, given demography information, our optimization tool determines at what power will broadband coverage will be feasible in the rural areas. This tool's design is separable and it can be utilized in other areas of the world, where demography and path-loss models can be replaced. Our optimization tool will be exemplified using results from one demographic region in India.

Channel codebook detection is of interest in cognitive paradigm or security applications. A binary hypothesis testing problem is considered, where a receiver has to detect the channel-code from two possible choices upon observing noise-affected codewords through a communication channel. For analytical tractability, it is assumed that the two channel-codes are linear block codes with identical block-length. In a first, this work studies the likelihood ratio test for minimizing the error probability in this detection problem. In an asymptotic setting, where a large number of noise-affected codewords are available for detection, the Chernoff information characterizes the error probability. A lower bound on the Chernoff information, based on the parameters of the two hypothesis, is established. Further, it is shown that if likelihood based efficient (generalized distributive law or BCJR) bit-decoding algorithms are available for the two codes, then the likelihood ratio test for the code-detection problem can be performed in a computationally feasible manner.

The sampling, quantization, and estimation of a bounded dynamic-range bandlimited signal affected by additive independent Gaussian noise is studied in this work. Considering the desirability of cheap, low-precision sensors, the use of single-bit analog to digital convertors (ADCs) is considered. For bandlimited signals, the distortion due to additive independent Gaussian noise can be reduced by oversampling (statistical diversity). The pointwise expected mean-squared error is used as a distortion metric for signal estimate in this work. If N is the oversampling ratio with respect to the Nyquist rate, then we show that a distortion of O(1=N) can be achieved with single-bit ADCs that record the signs of the observed noisy signal. This improves the (best known) distortion result by Masry for quantizing bandlimited signals in noise, using signs of noisy signal samples, from O(1/N2/3). This improvement comes by exploiting the structure of bandlimited signals in the estimation of original signal from noisy quantized bits.

SRAM failure-rate, induced by failure mechanisms such as process variations or cosmic neutrons, increases with reduction in the supply-voltage. Thus, appropriate supply-voltage margins must be used to mitigate or control the failure-rate. This work models and analyzes the interaction of two failure-mechanisms in an SRAM cell during the hold-operation (standby): (i) power-supply noise (due to switching at clock-edge) and soft-errors (due to radioactivity). Circuit-level simulations are used to analyze this interaction. The effect of random process-variations is accounted for by Monte-Carlo simulations. An error in stored-bit of SRAM cell is a transient phenomenon. Therefore, transient analysis of soft-error is carried out in the presence of power-supply noise and random process-variations by using circuit-level simulations in the UMC CMOS 90nm technology. A stochastic process based supply-voltage is difficult to model; therefore, a lower-envelope of measured supply-voltage from the literature is used as the noisy waveform. The critical-charge, obtained by using injection-current model of Freeman [1], of an SRAM cell is used to calculate the soft-error rate (SER). Simulation results show that the SER depends on the time-average of the noisy supply-voltage and not on its minimum value. The distribution of critical-charge, due to random process-variations, is estimated using simulations.

The problem of sampling a discrete-time sequence of spatially bandlimited fields, with a bounded dynamic range, in a distributed, communication-constrained processing environment is studied. A central unit having access to the data gathered by a dense network of low-precision sensors, is required to reconstruct the field snapshots to maximum accuracy. Both deterministic and stochastic field models are considered. For stochastic fields, results are established in the almost-sure sense. The feasibility of having a flexible tradeoff between the oversampling rate (sensor density) and the analog-to-digital converter (ADC) precision, while achieving an exponential accuracy in the number of bits per Nyquist-interval per snapshot is demonstrated. This exposes an underlying “conservation of bits” principle: the bit-budget per Nyquist-interval per snapshot (the rate) can be distributed along the amplitude axis (sensor-precision) and space (sensor density) in an almost arbitrary discrete-valued manner, while retaining the same (exponential) distortion-rate characteristics.

We study the problem of reducing power during data-retention in a standby static random access memory (SRAM). For successful data-retention, the supply voltage of an SRAM cell should be greater than a critical data retention voltage (DRV). Due to circuit parameter variations, the DRV for different cells on the same chip exhibits variation with a distribution having diminishing tail. For reliable data retention, the existing low-power design uses a worst-case technique in which a standby supply voltage that is larger than the highest DRV among all cells in an SRAM is used. Instead, our approach uses aggressive voltage reduction and counters the ensuing unreliability through a fault-tolerant memory architecture. The main results of this work are as follows: (i) We establish fundamental bounds on the power reduction in terms of the DRV-distribution using techniques from information theory. For the DRV-distribution of test-chip in (Qin, H, et al., 2006), we show that 49% power reduction with respect to (w.r.t.) the worst-case is a fundamental lower bound while 40% power reduction w.r.t. the worst-case is achievable with a practical combinatorial scheme, (ii) We study the power reduction as a function of the block-length for low-latency codes since most applications using SRAM are latency constrained. We propose a reliable memory architecture based on the Hamming code for the next test-chip implementation with a predicted power reduction of 33% while accounting for coding overheads.

The authors study leakage-power reduction in standby random access memories (SRAMs) during data-retention. An SRAM cell requires a minimum critical supply voltage (DRV) above which it preserves the stored-bit reliably. Due to process-variations, the intra-chip DRV exhibits variation with a distribution having a diminishing tail. In order to minimize leakage power while preserving data reliably, existing low-power design methods use a worst-case standby supply voltage. This worst-case voltage is larger than the highest DRV among all cells in an SRAM. In contrast, the approach uses aggressive voltage reduction and counters the ensuing unreliability by an error-control code based memory architecture. Using this approach, we explore fundamental trade-offs between power reduction and redundancy present in the SRAM. The authors establish fundamental bounds on the power reduction in terms of the DRV-distribution using techniques from information theory and algebraic coding theory. For an experimental test-chip DRV-distribution in the 90nm CMOS technology, the authors show that 49% power reduction with respect to (w.r.t.) the worst-case is a fundamental lower bound while 40% power reduction w.r.t. the worst-case is achievable by using a practical algebraic coding scheme. The authors also study the power reduction as a function of the block-length for low-latency codes since most applications using SRAM are latency constrained. The authors propose a reliable low-power memory architecture based on the Hamming code for the next test-chip implementation with a predicted power reduction of 33% while accounting for coding overheads

Distributed sampling and reconstruction of a physical field using an array of sensors is a problem of increasing interest in environmental monitoring applications of sensor networks. We show, using a dither-based scheme, that it is possible to reconstruct non-bandlimited fields with a reconstruction accuracy that depends on the available bitrate R and the spectral decay characteristics of the sensor field - we study exponentially decaying spectra as an illustration. For bandlimited fields f(t), the maximum pointwise error D/sub f/ decays as D/sub f/ /spl sim/ 2/sup -a1R/, i.e. exponentially with rate R. For the non-bandlimited case, we show that for fields u(t) with exponentially decaying spectral tails, i.e., |U(/spl omega/)|/spl sim/e/sup -a|/spl omega/|/, the maximum pointwise error D/sub u/ decays as D/sub u//spl sim/e/sup -a2/spl radic/R/(1+o(R)) with spatial bit rate R bits/metre. We also show that it is possible to trade off the number of sensors with their precision, while maintaining a similar reconstruction accuracy - a phenomenon that may be dubbed as a "bit-conservation" principle underlying the sampling framework.

Distributed sampling and reconstruction of a physical field using an array of sensors is a problem of considerable interest in environmental monitoring applications of sensor networks. Our recent work has focused on the sampling of bandlimited sensor fields. However, sensor fields are not perfectly bandlimited but typically have rapidly decaying spectra. In a classical sampling set-up it is possible to precede the A/D sampling operation with an appropriate analog anti-aliasing filter. However, in the case of sensor networks, this is infeasible since sampling must precede filtering. We show that even though the effects of aliasing on the reconstruction cannot be prevented due to the "filter-less" sampling constraint, they can be suitably controlled by oversampling and carefully reconstructing the field from the samples. We show using a dither-based scheme that it is possible to estimate non-bandlimited fields with a precision that depends on how fast the spectral content of the field decays. We develop a framework for analyzing non-bandlimited fields that lead to upper bounds on the maximum pointwise error for a spatial bit rate of R bits/meter. We present results for fields with exponentially decaying spectra as an illustration. In particular, we show that for fields f(t) with exponential tails; i.e., F(ω) ‹ παε–αω, the maximum pointwise error decays as c2e–α1√R+c3 1 over √R e >–2α1√R with spatial bit rate R bits/meter. Finally, we show that for fields with spectra that have a finite second moment, the distortion decreases as O((1 overN)2 over 3) as the density of sensors, N, scales up to infinity . We show that if D is the targeted non-zero distortion, then the required (finite) rate R scales as O (1 over √ overD log 1 over D).