We introduce GuessingGame, a protocol for evaluating large language models (LLMs) as strategic question-askers in open-ended, opendomain settings.A Guesser LLM identifies a hidden object by posing free-form questions to an Oracle without predefined choices or candidate lists.To measure question quality, we propose two information gain (IG) metrics: a Bayesian method that tracks belief updates over semantic concepts using LLM-scored relevance, and an entropy-based method that filters candidates via ConceptNet.Both metrics are model-agnostic and support post hoc analysis.Across 858 games with multiple models and prompting strategies, higher IG strongly predicts efficiency: a one-standard-deviation IG increase reduces expected game length by 43%.Prompting constraints guided by IG, such as enforcing question diversity, enable weaker models to significantly improve performance.These results show that question-asking in LLMs is both measurable and improvable, and crucial for interactive reasoning.

The softwarization and virtualization of network functions have enabled better flexibility and manageability in mobile communication systems today. However, considering the slowdown of Moore's law, it would be critical to reduce the cost of providing data plane services in the 6G era. In this paper, we argue that by leveraging programmable network devices such as Smart Network Interface Cards (SmartNICs) and programmable switches, we can offload custom network functionalities instead of using expensive host CPU resources for data plane operations. We propose S-Core, a scalable systems design for mobile packet core harnessing network data plane offloading. S-Core operates on top of multiple heterogeneous network devices, where heavy-hitter session rules can be dynamically placed in response to varying network traffic. Based on our prototype implementation, we argue that the use of programmable network devices will enable lessened CPU load, reduced latency, and improved energy efficiency in the next-gen communication systems.

Radio Access Network (RAN) data plane systems in 5G are expected to support multiple user traffic flows demanding high throughput and low latency. With increased usage of available services and applications, a single user will tend to have multiple streams of Internet Protocol (IP) flow and applications supported over a single Protocol Data Unit (PDU) session. Applications of similar Quality of Service (QoS) requirements get mapped to a single Data Radio Bearer (DRB) over the RAN. Several studies have found that over 90 % of total Internet traffic is Transmission Control Protocol (TCP) traffic. Single DRB can have multiple TCP flows in it. New Radio (NR) Packet Data Convergence Protocol (PDCP) layer delivers packets in-order to the upper layers for the configured DRB. When out-of-order packets are received, the reordering timer is triggered at PDCP and packets will wait in the reordering window until missing packets are received or the reorder timer expires. So a packet loss in one flow in the DRB can affect all the flows in the DRB and can lead to increased latency and reduced throughput in other TCP flows without loss. We propose a novel method to segregate TCP flows at a DRB in PDCP and handle data loss specific to each flow, so that flows without data loss are not impacted. When tested with multiple TCP flows under diverse loss conditions, the proposed solution yields a 4–9 % reduction in average latency and round-trip time (RTT) while boosting throughput when compared to NR PDCP.

As communication systems move towards 5G and beyond, utilizing a higher carrier frequency is essential to accommodate the exponential increment of data traffic. However, the short-range characteristics of high-frequency bands, such as mmWave and TeraHertz will cause frequent channel quality fluctuations. Also, radio protocol stack events triggered by these channel variations cause increased latency and reduced throughput. The existing transport protocols do not adapt well to such high variability of link quality and irregular network capacity, leading to the under-utilization of the network resources. More-over, the lack of radio link information hampers the transport layer in coping with dynamic channel conditions. The proposed Link Quality-assisted Transport (LQaT) solution estimates the link quality in real-time and notifies the end-host for adapting the transport layer to the varying wireless conditions. LQaT improves congestion/flow control to minimize the impact of link fluctuation and enhance the user experience. The experiments in ns-3 show that LQaT improves throughput up to 34% and reduces latency by up to 28% consistently, with effective link-layer adaptations.