Evidence and Data

This library of papers showcase peer-reviewed studies and scholarly articles that form the scientific backbone of our work. These publications detail rigorous experiments, analyses, and findings — providing transparency, context, and credibility. They allow experts, clinicians, and researchers to dive deep into the evidence behind our innovations, examine methodology, and build on our findings.

LLMs Will Always Hallucinate, and We Need to Live With This

1 February, 2024

The Vulnerability of Language Model Benchmarks: Do They Accurately Reflect True LLM Performance?

2 December, 2024

High-precision medical speech recognition through synthetic data and semantic correction: UNITED-MEDASR

24 November, 2024

First Train to Generate, then Generate to Train: UnitedSynT5 for Few-Shot NLI

12 December, 2024

Securing Well-Being: Exploring Security Protocols and Mitigating Risks in AI-Driven Mental Health Chatbots for Employees

1 January, 2024

Chatbot-Enhanced Mental Health First Aid in Corporate Settings: Addressing Risks, Implementing Crisis Management, and Promoting Employee Well-Being

10 November, 2024

Boosting Workplace Well-Being: A Novel Approach with a Mental Health Chatbot for Employee Engagement and Satisfaction

12 January, 2024

Our registered Patents highlight the unique inventions and technical solutions we’ve developed — from algorithms to processes — and for which we’ve secured formal intellectual-property protection. By listing our patents, we emphasise our commitment to original thinking and innovation.

Our World Records section celebrates standout achievements where we have reached — or surpassed — performance benchmarks on global leaderboards. These records reflect exceptional performance, often setting new standards in accuracy, efficiency or innovation.

Benchmark Type Score (asksam™) Prev. Record Prev. Holder Leaderboard
SNLI NLI 94.2% 93.1% EFL by Meta View SOTA
eSNLI NLI 94.01% 81.71% Oxford, UCL, Deepmind View SOTA
MultiNLI NLI 92.6% 91.8% OpenAI View SOTA
LibreSpeech ASR 0.986 (WER) 1.34 Google View SOTA
Europarl ASR 0.432 (WER) 7 MMPL View SOTA
Tedlium ASR 0.293 (WER) 3.92 NVIDIA View SOTA
FLEURS (English) ASR 0.336 (WER) 21.9 Meta (Multilingual) View SOTA

Curious how it works?

asksam does all that and more

01

Operates in a closed-source architecture

02

Integrates notes into a holistic case file

03

Provides patient specific outputs

04

Provides medically derived suggestions admin

05

Processes all types of medical documentation

06

Acts as your medical encyclopedia

Accesses trusted medical literature

Clinicians can asksam anything and as the platform is trained on medical literature within a closed-source environment, the platform can act as an on-demand encyclopaedia drawing from reputable medical literature without the risk of hallucination from the open-source internet.

In the premium tier, asksam also offers a Health AI Education Program that helps clinicians build digital literacy and understand how to safely incorporate AI tools into clinical workflows.

Provides patient specific outputs

asksam tailors summaries, letters, and explanations to reflect the patient’s specific medical history, demographics, and clinical needs.

As asksam operates in a closed-source environment there is no need to de-identify patient data, meaning reports and insights are specific to the patient not at a population level.

Provides medically derived suggestions

asksam processes all documentation loaded by the clinician and uses its Clinical Knowledge Graph to investigate all known medical associations to provide suggestions based on medical literature for the clinician to consider in their clinical decision making.

These notifications are designed purely to support administrative workflow and reduce cognitive load. They do not assess patient risk, monitor clinical parameters, or generate independent medical recommendations. 

Processes all types of medical data

asksam processes clinical documentation across PDF and Word Documents, pathology results, imaging summaries, and medication histories within the patient’s context.

Instead, asksam helps present information more clearly and coherently to support documentation workflows. It does not analyse or interpret medical data for diagnostic, predictive, or monitoring purpose.

Operates in a closed-sourced architecture

asksam is built as a fully closed-source clinical AI system, ensuring every component is tightly controlled, verified, and secured. This architecture prevents external data access, training leakage, or exposure to open internet sources.

All patient information stays within the clinical environment, protected by strict privacy and compliance guardrails. The result is a trusted, healthcare-exclusive AI platform clinicians can rely on.

Integrates notes into a holistic case file

asksam captures each consultation and automatically organises it into a structured, longitudinal case file. Notes, letters, templates, and follow-up plans are connected to the patient’s history, giving clinicians a complete view of their care.

This allows consistent documentation across episodes avoiding fragmented notes and the need to manually add context.