Publications

Selected publications

Cascading Eligibility Traces thumbnail
Learning From the Past with Cascading Eligibility Traces
Tokiniaina Raharison Ralambomihanta*, Ivan Anokhin*, Roman Pogodin*, Samira Ebrahimi Kahou, Jonathan Cornford, Blake Aaron Richards
ICLR 2026
[PDF, Code]
We show how to handle delayed credit assignment in an online, bio-plausible fashion by employing SSMs to retain past activations. We demonstrate the applicability of our method on computer vision and RL tasks. The design is friendly to on-device learning and Streaming RL.
Handling Delay in Real-Time RL thumbnail
Handling Delay in Real-Time Reinforcement Learning
Ivan Anokhin, Rishav Rishav, Matthew Riemer, Stephen Chung, Irina Rish, Samira Ebrahimi Kahou
ICLR 2025
[PDF, Code, Blog post, Poster]
We address real-time RL by pipelining layer computations and handling observation delay. We show it leads to faster inference and competitive performance on MuJoCo and MinAtar. The design is also friendly to on-device inference.

All publications

[1]
T. R. Ralambomihanta, I. Anokhin, R. Pogodin, S. E. Kahou, J. Cornford, and B. A. Richards, “Learning from the past with cascading eligibility traces,” arXiv preprint arXiv:2506.14598, 2025.
[2]
J. Chmura et al., “AIF-GEN: Open-source platform and synthetic dataset suite for reinforcement learning on large language models,” in Championing open-source DEvelopment in ML workshop@ ICML25,
[3]
I. Anokhin, R. Rishav, M. Riemer, S. Chung, I. Rish, and S. E. Kahou, “Handling delay in real-time reinforcement learning,” arXiv preprint arXiv:2503.23478, 2025.
[4]
I. Anokhin, R. Rishav, S. Chung, I. Rish, and S. E. Kahou, “Handling delay in reinforcement learning caused by parallel computations of neurons,” in ICML 2024 workshop: Aligning reinforcement learning experimentalists and theorists,
[5]
S. Chung, I. Anokhin, and D. Krueger, “Thinker: Learning to plan and act,” Advances in Neural Information Processing Systems, vol. 36, pp. 22896–22933, 2023.
[6]
M. Velikanov et al., “Embedded ensembles: Infinite width limit and operating regimes,” in International conference on artificial intelligence and statistics, PMLR, 2022, pp. 3138–3163.
[7]
I. Anokhin, K. Demochkin, T. Khakhulin, G. Sterkin, V. Lempitsky, and D. Korzhenkov, “Image generators with conditionally-independent pixel synthesis,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 14278–14287.
[8]
I. Anokhin et al., “High-resolution daytime translation without domain labels,” in Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2020, pp. 7488–7497.
[9]
I. Anokhin and D. Yarotsky, “Low-loss connection of weight vectors: Distribution-based approaches,” in International conference on machine learning, PMLR, 2020, pp. 335–344.