ADEPT Lab

A research lab at Brigham Young University building adaptive, person-centered methods for sequential decision-making in the messy, high-stakes settings where clean benchmarks run out.

Directed by Dr. Taylor W. Killian, Assistant Professor of Computer Science at Brigham Young University.

What we work on Joining the lab

About · What we do

Methods for decisions that have to be made.

We are less interested in clean problems with benchmarks than in the sequential, partially-observed ones that demand new thinking.

ADEPT builds adaptive, person-centered methods for real-world decision-making. We draw on reinforcement learning, uncertainty quantification, causal inference, and large language model reasoning, and we care most about the problems that separate genuine deployment from a leaderboard.

Those problems have a family resemblance. The state is only partially observed. The feedback that tells you whether a choice was good arrives late, or noisily, or not at all. The data was collected by someone else, for reasons you cannot fully reconstruct, which means it is confounded in ways no amount of fitting will undo. And some mistakes cannot be walked back. Acting well under that much genuine uncertainty is a methodological question, not an engineering detail.

Our work is grounded in clinical decision-making, where adaptive, context-sensitive choices have direct consequences for patients, and extends to scientific discovery. We are pragmatic by design: the aim is methods that researchers and practitioners can actually pick up and use.

Three square grid maps side by side, sharing a colour scale that runs from -1 in dark red through white at 0 to +1 in blue. The first, labelled LifeGate Env, shows the layout: a mostly black grid with a grey wall segment part-way down the left side, a yellow block filling the lower right, a blue strip along the top right edge, a red strip down the whole right edge, and a small white square marking the agent near the bottom. The second and third show the learned dead-end value and rescue value over those same cells. The dead-end map is pale in the upper left and deepens to dark red across the lower right; the rescue map is the near mirror image, deep blue through the upper left and fading to white toward the lower right.

In a small navigation task, separately learned dead-end and rescue values mark which regions of the grid still allow a good outcome and which no longer do.

  • Offline RL
  • Irreversibility & risk
  • Clinical & scientific

Medical Dead-ends and Learning to Identify High-Risk States and Treatments

A probability density curve, labelled Z-pi of s and a, drawn over return value on the horizontal axis and density on the vertical axis. The distribution has two humps: a small one over low returns near -0.5 and a much larger one near +0.2. The low-return tail beneath the small hump is shaded grey and cut by two vertical dashed lines — the value-at-risk threshold at roughly -0.5, and to its left the conditional value-at-risk, the average of that shaded tail, at roughly -0.65.

Conditional value at risk summarises the low-return tail of a return distribution instead of its average, which is what lets the method flag unrecoverable states at a chosen risk tolerance.

  • Offline RL
  • Irreversibility & risk
  • Clinical & scientific

Risk Sensitive Dead-end Identification in Safety-Critical Offline Reinforcement Learning

A left-to-right pipeline schematic. On the left, three dashed coloured traces of a multivariate time series X of t, with open circles marking observations at unevenly spaced times t0, t1, t2 through tT and the gap between two of them marked delta t. An arrow leads into a tapering block labelled Backbone NN, then a narrow block labelled Evidential Layer, which emits Normal Inverse-Wishart parameters phi. Two arrows branch to the right: the first to a shaded fan of predicted trajectories widening and narrowing around the observations, labelled calibrated, temporally correlated uncertainty; the second to a pair of small panels, a contour map of the distribution over the mean and covariance beside a scatter of orange and blue points split by a curved boundary, labelled uncertainty-guided, robust classification.

An irregularly sampled clinical time series is mapped to the parameters of an evidential distribution, so the model reports uncertainty that stays correlated across time rather than at isolated points.

  • Clinical & scientific

Continuous Time Evidential Distributions for Irregular Time Series

A tree diagram of a three-part action. From a filled root node, three branches labelled a1 = 0, a1 = 1 and a1 = 2 lead to boxed action vectors [0,0,0], [1,0,0] and [2,0,0], so each node holds a full action with the not-yet-chosen entries left at a default of zero. A branch labelled a2 = 1 descends from [0,0,0] to [0,1,0], which in turn splits along a3 = 1 and a3 = 2 to [0,1,1] and [0,1,2]. A smaller copy of the whole tree sits at lower right with one root-to-leaf path drawn in solid black and every other branch greyed out.

Laying a combinatorial action out as a tree lets the policy pick one sub-action per level and score a single path, rather than enumerating every combination of sub-actions.

  • Offline RL
  • Structured actions

BraVE: Offline Reinforcement Learning for Discrete Combinatorial Action Spaces

A left-to-right block diagram of a policy network. A circle holding the state s and a box labelled Sub-Action Embedding Table both feed a tall State Conditioning block, whose state-aware embeddings pass into a Transformer Block marked as repeated L times. Its output X enters a dashed group marked as repeated for each of the A sub-actions, containing a Decision MLP that emits distribution parameters and a Sample Sub-Action step that draws sub-action a-i. A Concat block gathers the sampled sub-actions into the final action vector a.

Self-attention over sub-action embeddings, conditioned on the state, lets each sub-action be chosen with the others in view and in no fixed order.

  • Structured actions

SAINT: Attention-Based Policies for Discrete Combinatorial Action Spaces

A line chart of learning progress against wall-clock time, with time in minutes from 0 to 300 on the horizontal axis and percentage of the reference FIQL asymptotic score on the vertical axis. A horizontal dashed line marks the 100 percent target. The thick dark navy SPIN curve sits flat at zero during pre-training, jumps past the target at a vertical dashed line annotated "ASM Learning Complete" around ten minutes, and stays above it. The three baseline curves — Factored IQL, SAINT and AR-IQL — climb gradually from the start and only approach the target somewhere between roughly 150 and 300 minutes.

On the humanoid-stand task, pre-training a model of which joint actions are valid lets policy learning reach the reference score within about ten minutes, where the baselines take hours.

  • Offline RL
  • Structured actions

Improving and Accelerating Offline RL in Large Discrete Action Spaces with Structured Policy Initialization

A three-part diagram summarising how to spend a fixed sampling budget, divided by dashed vertical rules and annotated with three numbered findings in salmon callout boxes. On the left, a set of three-dimensional axes — parallel rollouts per problem going up, problems per batch and sequential iterations along the base — carries a pale blue plane with dashed contours and a red line climbing through a row of yellow points, labelled: more compute, larger numbers of rollouts, on both easy and hard problem sets. In the middle, two small line charts plotted against sequential iterations: rollouts per problem rising in an S-shape, and problems per batch falling in the mirror image, labelled: with the total batch size held fixed, more iterations means more rollouts and fewer problems per batch. On the right, another set of the same three-dimensional axes with a flat orange surface low across the base, labelled: problems per batch has only a marginal effect within a moderate range.

A fixed sampling budget for reinforcement learning on language models is split three ways — rollouts per problem, problems per batch, and sequential iterations — and the number of rollouts per problem is the allocation that should grow as the budget grows.

IsoCompute Playbook: Optimally Scaling Sampling Compute for LLM RL

Lab · At a glance

5

Research areas

Five connected threads, running from offline reinforcement learning to the clinical and scientific settings.

2026

Lab launched

ADEPT opened in August 2026 within the BYU Department of Computer Science

Open

Recruiting now

Graduate and undergraduate researchers are being recruited for the lab’s first cohort.

Counts of students, publications, and funding will appear here as the lab accrues them.

Research · Areas

Offline RL

Learning to act from data someone else collected, for reasons you cannot fully reconstruct, in a world you cannot go back and query.

  • RL
  • Offline RL

Irreversibility & risk

States from which good outcomes are no longer reachable — how to recognize them, and how early you can be warned.

  • RL
  • Healthcare
  • Uncertainty

Delayed feedback

Credit assignment when outcomes arrive long after the decision that caused them — or arrive too corrupted to trust.

  • RL
  • Delayed Feedback
  • Scientific Discovery

Structured actions

Decisions composed of many interacting parts, where the number of joint actions is astronomically larger than anything you can enumerate.

  • RL
  • Offline RL

Clinical & scientific

The domains that generate the rest of our problems — where context-sensitive choices carry direct consequences.

  • Healthcare
  • Clinical
  • Scientific Discovery
All research areas

People · Principal investigator

Principal Investigator

People · Current students

Current Students

Avery Example

Offline policy learning when the logged data is thin.

  • PhD
  • RL
  • Offline RL
Email

Rosalind Placeholder

Works on treatment policies that have to be learned from observational records. Interested in what happens when the outcome that scores a decision does not arrive for weeks.

  • PhD
  • Healthcare
  • Delayed Feedback
  • Offline RL

Kai Sample

Uses language models to summarise clinical histories into states an agent can act on.

  • Master's
  • LLMs
  • Causal Inference
Email

Marguerite Testcase

Studies action spaces that are structured rather than flat — doses, timings and combinations that constrain each other. Building the evaluation setup the rest of the group borrows.

  • Master's
  • Structured Actions
  • RL

Devon Fixture

Runs the simulation sweeps and keeps the benchmark harness honest.

  • Undergraduate
  • RL
  • Healthcare

Tobias Mockup

Reads transcripts alongside the models that score them, looking for the cases where the two disagree.

  • Undergraduate
  • LLMs
  • Delayed Feedback

Open positions

The lab opened in August 2026 and is building its first cohort. Graduate and undergraduate researchers are being recruited now.

  • Prospective graduate students. Apply to the BYU CS graduate program and name ADEPT and the research areas that pull at you in your materials. Then email me at tkillian@cs.byu.edu to say you have applied.
  • BYU undergraduates. Email a short note about what you are excited about, have been building, or reading about.
  • Collaborators. Write about applied problems where the usual assumptions break.
Email tkillian@cs.byu.edu How to join
Full ADEPT Roster (past and present)

Publications · Featured

2026

Improving and Accelerating Offline RL in Large Discrete Action Spaces with Structured Policy Initialization

Matt Landers, Taylor W. Killian, Tom Hartvigsen, Afsaneh Doryab

ICLR 2026

Paper Code

IsoCompute Playbook: Optimally Scaling Sampling Compute for LLM RL

Zhoujun Cheng, Yutao Xie, Yuxiao Qu, Amrith Setlur, Shibo Hao, Varad Pimpalkhute, Tongtong Liang, Feng Yao, Zhengzhong Liu, Eric Xing, Virginia Smith, Ruslan Salakhutdinov, Zhiting Hu, Taylor W. Killian, Aviral Kumar

ICML 2026

Website Paper

2025

BraVE: Offline Reinforcement Learning for Discrete Combinatorial Action Spaces

Matt Landers, Taylor W. Killian, Hugo Barnes, Tom Hartvigsen, Afsaneh Doryab

NeurIPS 2025

Website Paper Code

SAINT: Attention-Based Policies for Discrete Combinatorial Action Spaces

Matt Landers, Taylor W. Killian, Tom Hartvigsen, Afsaneh Doryab

arXiv pre-print

Paper Code

2023

Risk Sensitive Dead-end Identification in Safety-Critical Offline Reinforcement Learning

Taylor W. Killian, Sonali Parbhoo, Marzyeh Ghassemi

Transactions on Machine Learning Research (TMLR)

Paper Forum Code

2021

Medical Dead-ends and Learning to Identify High-Risk States and Treatments

Mehdi Fatemi, Taylor W. Killian, Jayakumar Subramanian, Marzyeh Ghassemi

NeurIPS 2021

Paper Code MSR Blog
All publications

Lab · Recent news

What has been happening

  • The ADEPT Lab opens at BYU Computer Science. We are recruiting graduate and undergraduate researchers now.

  • The IsoCompute Playbook was accepted to ICML 2026.

  • SPIN, on offline RL in large discrete action spaces, was accepted to ICLR 2026.

All news

Joining · Now recruiting

Recruiting graduate and undergraduate researchers for the first cohort.

The next step for each group is on the joining page.

Email tkillian@cs.byu.edu How to join Research areas

Contact · Reach us

Get in touch

A short, specific email is always better than a long, general one.

Email

tkillian@cs.byu.edu

Where we are

Brigham Young University
Provo, Utah

Code and notes

github.com/byu-adept-lab