Benchmark scores are treated as evidence of LLM competence, but the same knowledge can yield different scores under different question formats. MCQs provide objective and scalable scoring, but their answer options can turn recall into recognition through answer-option scaffolding. Open-ended constructed response questions (CRQs) remove answer options, but their scores often depend on human or LLM judges and rubric interpretation, introducing judge-dependent scoring. This creates a trade-off between removing answer-option scaffolding and retaining objective scoring. We study whether this trade-off can be reduced through Constrained Structured Response Questions (CSRQs), a structured-response protocol that removes final answer options while preserving deterministic scoring. CSRQs use multi-select, ordering, and matching outputs as auditable answer spaces for exact verification. We evaluate CSRQs using converted knowledge benchmarks and MyCulture, which is, to our knowledge, the first Malaysia-centred cultural benchmark in Bahasa Malaysia, with parallel MCQ, CSRQ, and CRQ formats. Across benchmarks and model families, scores change when the evaluation instrument changes, i.e., scores drop after answer options are removed, and CRQ scores vary across judges. These findings suggest that benchmark scores should be interpreted as products of both model capability and the test used to measure it.

Aspect-Term Sentiment Analysis (ATSA) in multi-aspect sentences faces a fundamental tradeoff between efficiency and expressiveness. Existing models either re-encode the sentence for each aspect or rely on static use of deep representations, leading to redundant computation and limited adaptivity. We argue that Transformer depth is a costly, queryable resource, and propose DABS, a single-pass inference framework that encodes each sentence once to construct a reusable, depth-ordered substrate. Each aspect then queries this shared representation to selectively read relevant tokens and abstraction levels, without re-encoding. This decouples shared sentence encoding from lightweight, aspect-conditioned readout. Experiments on four ATSA benchmarks show that DABS achieves competitive performance while reducing end-to-end computation by up to 60% in multi-aspect settings (M ≥ 2). Further analyses indicate that adaptive depth querying is most beneficial for linguistically complex cases such as negation and contrast. Code is publicly available at https://github.com/panzhzh/acl-dabs.

This paper presents Ryt AI, an LLM-native agentic framework that powers Ryt Bank to enable customers to execute core financial transactions through natural language conversation. This represents the first global regulator-approved deployment worldwide where conversational AI functions as the primary banking interface, in contrast to prior assistants that have been limited to advisory or support roles. Built entirely in-house, Ryt AI is powered by ILMU, a closed-source LLM developed internally, and replaces rigid multi-screen workflows with a single dialogue orchestrated by four LLM-powered agents (Guardrails, Intent, Payment, and FAQ). Each agent attaches a task-specific LoRA adapter to ILMU, which is hosted within the bank’s infrastructure to ensure consistent behavior with minimal overhead. Deterministic guardrails, human-in-the-loop confirmation, and a stateless audit architecture provide defense-in-depth for security and compliance. The result is Banking Done Right: demonstrating that regulator-approved natural-language interfaces can reliably support core financial operations under strict governance.

Capitalise on deep learning models, offering Natural Language Processing (NLP) solutions as a part of the Machine Learning as a Service (MLaaS) has generated handsome revenues. At the same time, it is known that the creation of these lucrative deep models is non-trivial. Therefore, protecting these inventions’ intellectual property rights (IPR) from being abused, stolen and plagiarized is vital. This paper proposes a practical approach for the IPR protection on recurrent neural networks (RNN) without all the bells and whistles of existing IPR solutions. Particularly, we introduce the Gatekeeper concept that resembles the recurrent nature in RNN architecture to embed keys. Also, we design the model training scheme in a way such that the protected RNN model will retain its original performance iff a genuine key is presented. Extensive experiments showed that our protection scheme is robust and effective against ambiguity and removal attacks in both white-box and black-box protection schemes on different RNN variants. Code is available at https://github.com/zhiqin1998/RecurrentIPR.