Postdoctoral Fellow - Uncertainty Quantification and Calibration for Large Language Models

The program

The NRC’s Postdoctoral Fellowship program offers PDFs access to unique world-class facilities and the opportunity to work alongside multi-disciplinary teams of expert researchers and technicians on projects of critical importance to Canada. PDFs will carry out research on innovative research projects, with opportunities for career development (publications and/or industry interaction).

PDFs will be offered appointments to the staff of the National Research Council on a term basis and will be offered salaries and benefits currently available to Research Officers. Salaries commensurate with experience are taxable and subject to other deductions. As a guide, the current annual PhD recruiting rate is $83,985. In addition PDFs will be allocated a travel fund to support their work.

Project title

Uncertainty Quantification and Calibration for Large Language Models

The role

Are you looking for an opportunity to conduct cutting-edge research with real-life impacts? Do you want to grow in a team environment that provides guidance and autonomy? Do you want ownership over outcomes allowing you to present your work at competitive international venues? Look no further we have a rewarding opportunity for you!

Current human-AI interaction is plagued by over-reliance on AI predictions, an inability to distinguish AI-generated content, and the risks of AI-generated falsehoods and deceptions. Quantifying the uncertainty in AI generations is a pathway to mitigating all of these risks.

Traditionally in AI research, a well-designed certainty measure assigns a probability of correctness to each generated output. Combining AI outputs with these interpretable confidence scores is immensely valuable for informed human-AI interaction. In the new era of generative large language models (LLMs), uncertainty quantification becomes less well-defined due to the open-ended nature of LLM outputs (i.e., it is necessary to simultaneously estimate the uncertainty in what is being said, and how it is being said, while dealing with an infinite output space). However, LLMs are now the backbone of most human-facing AI systems. How to meaningfully capture these uncertainties to provide useful context on LLM generations is an active research question.

Separately, recent work suggests that LLMs represent human-interpretable concepts (e.g., emotions, truthfulness, harmfulness, etc.) at intermediate points in their computation. This opens up a new opportunity to inspect uncertainty in the LLMs’ encoding of these concepts, which would enable new insights and intervention mechanisms. To date, we are not aware of any work that applies certainty calibration methods to intermediate model representations.

This project develops methods for uncertainty quantification that can be applied to both output generations, and intermediate features. Foundational work on this subject is both technically challenging and highly applicable to a wide range of open problems. The anticipated research outcomes are appropriate for publication in top-tier international venues for Computational Linguistics and Machine Learning research, such as ACL* conferences, ICLR, or ICML.

The Research Problem

Large language models produce text by iteratively generating a probability distribution over possible next-word choices, then sampling from that distribution. The problem of uncertainty quantification aims to consolidate that probability distribution into a few meaningful and interpretable certainty metrics. Certainty metrics can answer questions about reproducibility (i.e., if asked the same thing again, how likely is the LLM to respond in the same way?), or embedded knowledge (i.e., if asked a factual query, how certain is the LLM about its prediction?). Such measures can help contextualize LLM outputs for users or developers, and are a critical ingredient for trustworthy AI systems.

The proposed research agenda develops new methods for uncertainty quantification that can apply to features, determines whether such signals can be used to improve uncertainty metrics for text generation, and enables a host of impactful research applications:

• Hallucination detection
• Deception detection
• Synthetic content detection 
• Mitigation of safety failure modes  
• Improved evaluation  
• Trustworthy decision support systems 

In the long-term, the new knowledge that is created can support Canada’s commitment to safe and responsible use of generative AI, support strategic defence applications, and open new research opportunities.

In addition to the proposed research program, you will benefit from proximity to related work that is ongoing in the team. We currently work on distinct, but related, work in LLM interpretability and evaluation. There is potential for you to contribute to those works as well, if desired. We have ongoing collaborations with Mila (one of Canada’s leading AI research institutes) and Cohere (Canada’s largest LLM provider), which creates good   networking or collaboration opportunities. It is expected that you will work autonomously, with guidance from the project lead, such that you have ownership over the outcomes, obtain first-author publications, and present the work at these competitive international venues.

Education

A  Ph.D. in Computer Science, Mathematics, Statistics, Physics, or related discipline is required. 

For information on certificates and diplomas issued abroad, please see Degree equivalency

Experience

• Experience with the design and implementation of novel and speculative research projects in AI-related fields.
• Experience with deep learning tools and libraries, such as those specialized for data collection and analysis, or data simulation. 
• Experienced publication record with evidence of significant contribution in experimental design and implementation, scientific writing and presentation at high-calibre venues.
• Experience with leading reading groups, mentoring students, or organizing workshops is an asset.
• Experience understanding uncertainty quantification methods or mechanistic interpretability methods is an asset.

Language requirements

English

Information on language requirements and self-assessment tests

Condition of employment

Reliability Status

For a Reliability Status, verification of background information over a period of 5 years is required.

Who is eligible?

• Fellowships will be for two years contingent on satisfactory progress achieved during the first year.
• Candidates should have obtained a PhD (or equivalent) within the past three years (PhD received on or after July 1, 2024*) or expect to complete their PhD within 6 months of appointment.
• Fellows will work in a lab under the direct supervision of an NRC researcher.

* Exceptions may be considered to recognize personal or family related leave. Please connect with us using the contact information below to discuss in more detail.

Application requirements

To be considered for the program please include the following in your application, please note that you will need to attach the required documents as per the list below when submitting your application. Failure to do so will result in your application being excluded.

• Resume
• Statement of Interest in the project (maximum one page in length)
• PhD Transcript - an electronic copy is sufficient, it does not have to be an official version.
• List of Publications

When submitting your application you can include the required documents in any attachment field such as «Second language evaluation results» or «Other attachments»

In addition, applicants who best meet the requirements of the position will be asked to provide three letters of recommendation at a later stage of the competition process.

Compensation

The intent of this hiring action is to staff the position at the AsRO level, which is an early-career level position with a salary range of $83,985 to $116,641. This wide range accounts for varying degrees of expertise and impact, and growth within the role.

Salary determination will be based on a thorough evaluation of the successful candidate's prior experience. This evaluation will consider specific achievements, skills, and contributions that demonstrate the candidate's impact and expertise. As a guide, the average starting rate for new PhD graduates entering this role is $83,985. 

The NRC is committed to supporting career progression within the organization, providing access to the opportunities and resources needed for a pathway to a fulfilling career. NRC employees also enjoy a wide-range of competitive benefits including a robust pension plan, comprehensive health and dental coverage, disability and life insurance, office closure at the end of December, and additional supports to enhance your well-being throughout your career and beyond.

Discover the possible

Anything is possible at the NRC, named in 2026 one of Canada’s Top Employers for Young People, Top Employer in the National Capital Region and Forbes Canada’s Best Employers!

As Canada’s largest research and innovation organization, our world-renowned research pushes the boundaries of science and engineering to make the impossible, possible. Every day we explore new ideas through innovative research and help companies discover possibilities that impact Canada’s future and the world.

At the NRC, you’ll also discover new possibilities. Our supportive workplace fosters a culture of creativity, welcoming fresh perspectives and innovation at all levels. We value teamwork. You’ll collaborate across multiple fields and with the brightest minds to find creative solutions. Most importantly, you’ll discover what’s possible within you as you grow, make valuable contributions and progress in your professional journey. From ground-breaking discoveries to a life-changing career, discover your possible at the NRC.

Notes

• In 2026, the NRC was chosen as one of Canada’s Top Employers for Young People, a National Capital Region Top Employer and Forbes Canada’s Best Employer.
• Relocation assistance will be determined in accordance with the NRC's directives.
• Fellowships are open to nationals of all countries, although preference will be given to Canadian citizens and permanent residents of Canada. Please include citizenship information in your application.
• The incumbent must adhere to safe workplace practices at all times.
• We thank all those who apply, however only those selected for further consideration will be contacted.