Skip to content

Commit 17143b8

Browse files
authored
Merge pull request #299 from HIRO-group/stephane_final
add paper pdfs and bibs, add thesis, add news about papers and thesis
2 parents d680082 + 929aef6 commit 17143b8

9 files changed

+87
-1
lines changed

_data/people.yml

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -213,7 +213,7 @@
213213
role: alumni-spotlight
214214
url: https://stephao.github.io/
215215
graduated: HIRO Ph.D. [2020-2025]
216-
nowat: Senior ML Research Scientist at Skyfall AI
216+
nowat: Research Scientist at Skyfall AI
217217
pic: stephane.png
218218

219219
- name: Sandra Bae

_data/publications.yml

Lines changed: 15 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,10 @@
1+
- title: "ReSeeding Latent States for Sequential Language Understanding"
2+
authors: Stéphane Aroca-Ouellette, Katharina von der Wense, Alessandro Roncone
3+
year: 2025
4+
venue: "Conference on Empirical Methods in Natural Language Processing [EMNLP]"
5+
where: Suzhou, China, November 1-5
6+
pdf: "2025_Aroca-Ouellette_EMNLP"
7+
subteams: [learning and modeling, stephane, conference]
18
- title: "Dynamics-Compliant Trajectory Diffusion for Super-Nominal Payload Manipulation"
29
authors: Anuj Pasricha, Joewie Koh, Jay Vakil, Alessandro Roncone
310
year: 2025
@@ -18,6 +25,14 @@
1825
where: Montreal, Canada, August 16-22
1926
pdf: "2025_Aroca-Ouellette_IJCAI_HA2"
2027
subteams: [algorithmic hri and human-ai teaming, stephane, conference, social intelligence]
28+
- title: "Toward Human-Inspired AI: Identifying Data, Building Structures, and Hypothesis-Driven Learning"
29+
authors: Stéphane Aroca-Ouellette
30+
year: 2025
31+
venue: University of Colorado Boulder
32+
where: Boulder, CO, U.S.A.
33+
description: PhD dissertation
34+
subteams: [learning and modeling, stephane, thesis, social intelligence]
35+
pdf: "2025_Aroca-Ouellette_Thesis"
2136
- title: "GenTact Toolbox: A Computational Design Pipeline to Procedurally Generate Context-Driven 3D Printed Whole-Body Artificial Skins"
2237
authors: Carson Kohlbrenner, Caleb Escobedo, S. Sandra Bae, Alexander Dickhans, Alessandro Roncone
2338
year: 2025

_posts/news/2025-08-22-IJCAI.md

Lines changed: 12 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,12 @@
1+
---
2+
title: "📑 Paper on Hierarchical Reinforcement Learning for Human-Agent Interaction published at IJCAI 2025!"
3+
description: Implicitly Aligning Humans and Autonomous Agents through Shared Task Abstractions
4+
tags: [news]
5+
author: Stéphane Aroca-Ouellette
6+
---
7+
8+
Our paper: "Implicitly Aligning Humans and Autonomous Agents through Shared Task Abstractions" was recently published to IJCAI 2025!
9+
10+
This work introduces a new method (HA²) that enables AI systems to use human-like task structures. Empirical results, including a human study, show that this induced structures allows AI teammates to coordinate more effectively with unfamiliar partners—both other AI systems and real people—and outperform all existing approaches.
11+
12+
[[PDF]]({{ site.url }}/papers/2025_Aroca-Ouellette_IJCAI_HA2.pdf) [[BIB]]({{ site.url }}/papers/2025_Aroca-Ouellette_IJCAI_HA2.bib)

_posts/news/2025-11-05-EMNLP.md

Lines changed: 12 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,12 @@
1+
---
2+
title: "📑 Paper on Grounding Language for Sequential Resaoning published at EMNLP 2025!"
3+
description: ReSeeding Latent States for Sequential Language Understanding
4+
tags: [news]
5+
author: Stéphane Aroca-Ouellette
6+
---
7+
8+
Our paper: "ReSeeding Latent States for Sequential Language Understanding" was recently published to EMNLP 2025!
9+
10+
This work introduces ReSEED, a method that helps AI models better understand and reason about sequences by grounding their language representations in real environmental data. Experiments on new benchmarks show that ReSEED generalizes much better than text-only models and even outperforms commercial LLMs, demonstrating the value of linking language to the actual state of the world.
11+
12+
[[PDF]]({{ site.url }}/papers/2025_Aroca-Ouellette_EMNLP.pdf) [[BIB]]({{ site.url }}/papers/2025_Aroca-Ouellette_EMNLP.bib)
Lines changed: 14 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,14 @@
1+
---
2+
title: "📢 Stéphane's successful PhD dissertation defense!!"
3+
description:
4+
tags: [news]
5+
author: Alessandro Roncone
6+
---
7+
8+
We are thrilled to congratulate *Dr. Stéphane Aroca-Ouellette*{:.color-banner}, who successfully defended his PhD dissertation titled _"Toward Human-Inspired AI: Identifying Data, Building Structures, and Hypothesis-Driven Learning"_. Stéphane's hard work has paid off, and the dissertation committee (composed of Alessandro Roncone, Katharina von der Wense, Bradley Hayes, Maria Pacheco, and Rin Metcalf Susa) was happy to see his continued maturation toward research independence. Congratulations Dr. Aroca-Ouellette!
9+
10+
His thesis PDF is available [[here]]({{ site.url }}papers/2025_Aroca-Ouellette_Defense.pdf), and the defense presentation can be viewed below:
11+
12+
{% include video.html url="//www.youtube.com/embed/IS3G2uOjszM" start="102" margin-bottom="40px" padding-bottom="52%" max-width="100%" %}
13+
14+
Stéphane has started his new position as <ins>Research Scientist at Skyfall AI</ins>---we wish him every success in his future endeavors!
Lines changed: 20 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,20 @@
1+
@inproceedings{aroca-ouellette-etal-2025-reseeding,
2+
title = "{R}e{S}eeding Latent States for Sequential Language Understanding",
3+
author = "Aroca-Ouellette, St{\'e}phane and
4+
von der Wense, Katharina and
5+
Roncone, Alessandro",
6+
editor = "Christodoulopoulos, Christos and
7+
Chakraborty, Tanmoy and
8+
Rose, Carolyn and
9+
Peng, Violet",
10+
booktitle = "Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing",
11+
month = nov,
12+
year = "2025",
13+
address = "Suzhou, China",
14+
publisher = "Association for Computational Linguistics",
15+
url = "https://aclanthology.org/2025.emnlp-main.1281/",
16+
doi = "10.18653/v1/2025.emnlp-main.1281",
17+
pages = "25233--25247",
18+
ISBN = "979-8-89176-332-6",
19+
abstract = "We introduce Refeeding State Embeddings aligned using Environmental Data (ReSEED), a novel method for grounding language in environmental data. While large language models (LLMs) excel at many tasks, they continue to struggle with multi-step sequential reasoning. ReSEED addresses this by producing latent embeddings aligned with the true state of the environment and refeeding these embeddings into the model before generating its output. To evaluate its effectiveness, we develop three new sequential reasoning benchmarks, each with a training set of paired state-text trajectories and several text-only evaluation sets that test generalization to longer, unseen trajectories. Across all benchmarks, ReSEED significantly improves generalization and scalability over a text-only baseline. We further show that ReSEED outperforms commercial LLMs on our benchmarks, highlighting the value of grounding language in the environment."
20+
}
714 KB
Binary file not shown.
Lines changed: 13 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,13 @@
1+
@phdthesis{aroca-ouellette2025humaninspired,
2+
title = {Toward Human-Inspired AI: Identifying Data, Building Structures, and Hypothesis-Driven Learning},
3+
author = {Aroca-Ouellette, Stephane},
4+
year = {2025},
5+
month = dec,
6+
school = {University of Colorado Boulder},
7+
address = {Boulder, CO, USA},
8+
doi = {10.25346/S6/32245644}, % If ProQuest provides an official DOI, replace this placeholder
9+
url = {https://www.proquest.com/docview/32245644}, % Replace if you have the final repository link
10+
type = {Doctor of Philosophy Dissertation},
11+
orcid = {0000-0001-5880-8430},
12+
abstract = {Artificial intelligence (AI) systems frequently fail to generalize effectively to novel and out-of-distribution scenarios, fundamentally due to their sole reliance on extensive, data-driven learning. Such methods are known to exploit surface-level correlations rather than learn deeper causal structures, leading to significant issues in scenarios marked by data scarcity or novelty. In contrast, humans excel in generalization through causal reasoning, efficiently adapting existing knowledge and continuously refining it through hypothesis-driven learning. This thesis investigates how core human cognitive mechanisms—specifically data selection, structural abstraction, and hypothesis-driven learning—can inspire algorithmic advancements to address AI's generalization limitations. First, we demonstrate the importance of comprehensive multi-modal data streams, showing that richer, contextually grounded data enhances generalization in natural language understanding and human-agent collaboration. Next, we explore structured representations by proposing a hierarchical reinforcement learning framework mirroring human cognitive structures, significantly improving agent adaptability in human-agent collaboration. Finally, we introduce PROSE, a hypothesis-driven learning method enabling AI models to rapidly infer and iteratively refine latent user preferences from limited data. Collectively, this thesis underscores the potential of human-inspired methodologies to create AI systems that not only generalize more robustly but are inherently better aligned with human norms and expectations, paving the way toward truly adaptive, human-centered artificial intelligence.}
13+
}
6.54 MB
Binary file not shown.

0 commit comments

Comments
 (0)