arXiv subjects

Atul Anand

Publications and source records attributed to Atul Anand.

4 recordsLinked to original sources

arXiv2026

Diagnosing Tool-Selection Reasoning in LLM Agents with Canary Tools

Agent evaluations tell us that a model picked the wrong tool, but rarely why. We introduce canary tools: diagnostic probe tools planted in an agent's Model Context Protocol (MCP) tool set, each engineered to probe one specific tool-selection weakness. A six-type taxonomy (semantic decoys, parameter traps, capability mirages, prerequisite blindness, temporal decoys, and granularity traps) turns a single "wrong tool" outcome into a multi-dimensional profile of how a model reasons about tools. We evaluate eight models -- six hosted and two 8B open-weight -- spanning three capability tiers, on 120 tasks across three canary-density conditions and three seeds (8,640 runs), plus a 2,880-run subtlety ablation. Task success is graded by a provider-independent judge, corroborated by a second independent judge (Cohen's kappa = 0.75). We report three findings. First, susceptibility drops sharply as models get more capable: the per-task canary susceptibility rate (CSR) ranges about 36x across models, lowest for Claude Opus 4.8 and highest for Llama 3.1 8B. Second, capability tier alone does not predict safety: the most susceptible hosted model is mid-tier, and within a provider the cheaper model can be the safer one. Third, the taxonomy is capability-stratified: capability mirages most reliably trap frontier models, while the other types are largely inert on strong models but fire on small open models, so they discriminate by capability rather than being weak. Softening each canary's give-away phrase leaves frontier CSR essentially unchanged, evidence that the probes measure reasoning, not phrase-spotting. Susceptibility also predicts task failure (Spearman rho = -0.34), while the most robust models are not significantly degraded by canary pressure. We release the framework, canary schemas, tasks, and logs.

cs.AI↗

arXiv2026

Instruction Stacking Collapse: A Benchmark and the Capability-Dependent Value of Prompt Compilation

Production prompts rarely carry a single instruction. One system message may require valid JSON, a word limit, three citations, and a fixed tone at the same time. We study how instruction-following degrades as such constraints accumulate. We introduce a benchmark that stacks 24 verifier-checked instructions, one to twenty at a time, and evaluate three production-tier LLMs (Claude Sonnet 4.6, GPT-5-mini, Gemini 2.5 Flash). Instruction-following degrades non-linearly: the follow rate falls from ~96% to as low as 20%, driven by a structured and reproducible set of pairwise conflicts. A single "output JSON" constraint, for example, is jointly unsatisfiable with nine others. We then evaluate a training-free remedy: an instruction compiler that rewrites the stacked prompt in a single LLM call and is reused across queries. Its benefit is capability-graded. It recovers up to +11 points of follow rate for weaker models, which are also the models most often deployed at scale, while leaving stronger models, which already internalise the same structure, essentially unchanged. Cluster-robust tests, same-baseline controls, and a within-family scaling ladder attribute the gain to the rewrite itself rather than to additional tokens, reordering, or measurement headroom. We release the benchmark, verifiers, and cached runs for full reproduction.

cs.SE↗

arXiv2025

Adverse Event Extraction from Discharge Summaries: A New Dataset, Annotation Scheme, and Initial Findings

In this work, we present a manually annotated corpus for Adverse Event (AE) extraction from discharge summaries of elderly patients, a population often underrepresented in clinical NLP resources. The dataset includes 14 clinically significant AEs-such as falls, delirium, and intracranial haemorrhage, along with contextual attributes like negation, diagnosis type, and in-hospital occurrence. Uniquely, the annotation schema supports both discontinuous and overlapping entities, addressing challenges rarely tackled in prior work. We evaluate multiple models using FlairNLP across three annotation granularities: fine-grained, coarse-grained, and coarse-grained with negation. While transformer-based models (e.g., BERT-cased) achieve strong performance on document-level coarse-grained extraction (F1 = 0.943), performance drops notably for fine-grained entity-level tasks (e.g., F1 = 0.675), particularly for rare events and complex attributes. These results demonstrate that despite high-level scores, significant challenges remain in detecting underrepresented AEs and capturing nuanced clinical language. Developed within a Trusted Research Environment (TRE), the dataset is available upon request via DataLoch and serves as a robust benchmark for evaluating AE extraction methods and supporting future cross-dataset generalisation.

cs.CL↗

arXiv2024

Frailty or Frailties: Exploring Frailty Index Subdimensions in the English Longitudinal Study of Ageing

Background: Frailty, a state of increased vulnerability to adverse health outcomes, has garnered significant attention in research and clinical practice. Existing constructs aggregate clinical features or health deficits into a single score. While simple and interpretable, this approach may overlook the complexity of frailty and not capture the full range of variation between individuals. Methods: Exploratory factor analysis was used to infer latent dimensions of a frailty index constructed using survey data from the English Longitudinal Study of Ageing (ELSA), wave 9. The dataset included 58 self-reported health deficits in a representative sample of community-dwelling adults aged 65+ (N = 4971). Deficits encompassed chronic disease, general health status, mobility, independence with activities of daily living, psychological wellbeing, memory and cognition. Multiple linear regression examined associations with CASP-19 quality of life scores. Results: Factor analysis revealed four frailty subdimensions. Based on the component deficits with the highest loading values, these factors were labelled "Mobility Impairment and Physical Morbidity", "Difficulties in Daily Activities", "Mental Health" and "Disorientation in Time". The four subdimensions were a better predictor of quality of life than frailty index scores. Conclusions: Distinct subdimensions of frailty can be identified from standard index scores. A decomposed approach to understanding frailty has potential to provide a more nuanced understanding of an individual's state of health across multiple deficits.

stat.AP↗