Banafsheh Rafiee

Banafsheh Rafiee, Sina Ghiassian, Jun Jin et al. · deepmind

In this paper, we explore an approach to auxiliary task discovery in reinforcement learning based on ideas from representation learning. Auxiliary tasks tend to improve data efficiency by forcing the agent to learn auxiliary prediction and control objectives in addition to the main task of maximizing reward, and thus producing better representations. Typically these tasks are designed by people. Meta-learning offers a promising avenue for automatic task discovery; however, these methods are computationally expensive and challenging to tune in practice. In this paper, we explore a complementary approach to the auxiliary task discovery: continually generating new auxiliary tasks and preserving only those with high utility. We also introduce a new measure of auxiliary tasks' usefulness based on how useful the features induced by them are for the main task. Our discovery algorithm significantly outperforms random tasks and learning without auxiliary tasks across a suite of environments.

Banafsheh Rafiee, Jun Jin, Jun Luo et al. · deepmind

Auxiliary tasks have been argued to be useful for representation learning in reinforcement learning. Although many auxiliary tasks have been empirically shown to be effective for accelerating learning on the main task, it is not yet clear what makes useful auxiliary tasks. Some of the most promising results are on the pixel control, reward prediction, and the next state prediction auxiliary tasks; however, the empirical results are mixed, showing substantial improvements in some cases and marginal improvements in others. Careful investigations of how auxiliary tasks help the learning of the main task is necessary. In this paper, we take a step studying the effect of the target policies on the usefulness of the auxiliary tasks formulated as general value functions. General value functions consist of three core elements: 1) policy 2) cumulant 3) continuation function. Our focus on the role of the target policy of the auxiliary tasks is motivated by the fact that the target policy determines the behavior about which the agent wants to make a prediction and the state-action distribution that the agent is trained on, which further affects the main task learning. Our study provides insights about questions such as: Does a greedy policy result in bigger improvement gains compared to other policies? Is it best to set the auxiliary task policy to be the same as the main task policy? Does the choice of the target policy have a substantial effect on the achieved performance gain or simple strategies for setting the policy, such as using a uniformly random policy, work as well? Our empirical results suggest that: 1) Auxiliary tasks with the greedy policy tend to be useful. 2) Most policies, including a uniformly random policy, tend to improve over the baseline. 3) Surprisingly, the main task policy tends to be less useful compared to other policies.

Banafsheh Rafiee, Richard Sutton

In this paper, we advocate for incorporating enactive approaches to perception and cognition into artificial intelligence (AI). Enactive approaches view perception as an active, skillful engagement with the world, where agents perceive by acting and by understanding how their actions shape their experience. This contrasts with classical views that treat perception as a passive internal process in which the brain receives sensory input, processes it, and issues commands for action. Enactive views emphasize the dynamic, embodied, and interactive character of perception, grounded in the lived experience of agents embedded in their environments. We identify and develop four key enactive concepts that we find most relevant to AI: experience, action perception inseparability, autonomy, and embodiment. Much of mainstream AI, from classical rule based systems to large language models, has largely neglected these insights, treating cognition as internal processing detached from embodied interaction and intrinsic normativity. Reinforcement learning (RL), however, exhibits structural resonance with enactive principles through its emphasis on action, agent environment interaction, feedback driven adaptation, and agent centered evaluation. However, this resonance should not be taken as theoretical equivalence, as RL approximates some enactive insights, but key elements remain absent or weakly developed. Building on this analysis, we suggest a broader incorporation of enactive ideas into both mainstream AI and RL.

Kamil Ciosek, Banafsheh Rafiee, Sina Ghiassian et al.

We characterize the minimax rate of estimating the second-order calibration error for binary classification, which quantifies whether a higher-order predictor's epistemic-uncertainty estimate matches the conditional variance of the label probability on its level sets. Our key observation is that the sech perturbation kernel, previously used only to enforce smoothness of calibration functions, in fact makes them analytic in a strip of half-width $hπ/2$. Polynomial regression then estimates the calibration error at rate $\tilde{O}(1/\sqrt{n})$, with explicit constants, a qualitative improvement over the $O(n^{-1/4})$ rate achievable by bucketing or kernel smoothing. A matching $Ω(1/\sqrt{n})$ lower bound establishes minimax optimality up to logarithmic factors. As a corollary, we give the first finite-sample guarantee for second-order Platt scaling, yielding a post-hoc procedure that recalibrates both the mean prediction and the epistemic-variance estimate of any higher-order predictor. Along the way, we provide a bucket-free definition of second-order calibration and relate it quantitatively to the bucketed formulation of Ahdritz et al. [2025]. Our experiments confirm the predicted rate and the quality of the recalibrated uncertainties.

Mehdi Fatemi, Banafsheh Rafiee, Mingjie Tang et al. · ibm-research

Despite significant advancements in large language models (LLMs), a major drawback of reasoning models is their enormous token usage, which increases computational cost, resource requirements, and response time. In this work, we revisit the core principles of reinforcement learning (RL) and, through mathematical analysis, demonstrate that the tendency to generate lengthy responses arises inherently from RL-based optimization during training. This finding questions the prevailing assumption that longer responses inherently improve reasoning accuracy. Instead, we uncover a natural correlation between conciseness and accuracy that has been largely overlooked. We show that introducing a secondary phase of RL training, using a very small set of problems, can significantly reduce chains of thought while maintaining or even enhancing accuracy. Additionally, we demonstrate that, while GRPO shares some interesting properties of PPO, it suffers from collapse modes, which limit its reliability for concise reasoning. Finally, we validate our conclusions through extensive experimental results.

Banafsheh Rafiee, Zaheer Abbas, Sina Ghiassian et al.

We present three new diagnostic prediction problems inspired by classical-conditioning experiments to facilitate research in online prediction learning. Experiments in classical conditioning show that animals such as rabbits, pigeons, and dogs can make long temporal associations that enable multi-step prediction. To replicate this remarkable ability, an agent must construct an internal state representation that summarizes its interaction history. Recurrent neural networks can automatically construct state and learn temporal associations. However, the current training methods are prohibitively expensive for online prediction -- continual learning on every time step -- which is the focus of this paper. Our proposed problems test the learning capabilities that animals readily exhibit and highlight the limitations of the current recurrent learning methods. While the proposed problems are nontrivial, they are still amenable to extensive testing and analysis in the small-compute regime, thereby enabling researchers to study issues in isolation, ultimately accelerating progress towards scalable online representation learning methods.

Sina Ghiassian, Banafsheh Rafiee, Yat Long Lo et al.

Reinforcement learning systems require good representations to work well. For decades practical success in reinforcement learning was limited to small domains. Deep reinforcement learning systems, on the other hand, are scalable, not dependent on domain specific prior knowledge and have been successfully used to play Atari, in 3D navigation from pixels, and to control high degree of freedom robots. Unfortunately, the performance of deep reinforcement learning systems is sensitive to hyper-parameter settings and architecture choices. Even well tuned systems exhibit significant instability both within a trial and across experiment replications. In practice, significant expertise and trial and error are usually required to achieve good performance. One potential source of the problem is known as catastrophic interference: when later training decreases performance by overriding previous learning. Interestingly, the powerful generalization that makes Neural Networks (NN) so effective in batch supervised learning might explain the challenges when applying them in reinforcement learning tasks. In this paper, we explore how online NN training and interference interact in reinforcement learning. We find that simply re-mapping the input observations to a high-dimensional space improves learning speed and parameter sensitivity. We also show this preprocessing reduces interference in prediction tasks. More practically, we provide a simple approach to NN training that is easy to implement, and requires little additional computation. We demonstrate that our approach improves performance in both prediction and control with an extensive batch of experiments in classic control domains.

Sina Ghiassian, Huizhen Yu, Banafsheh Rafiee et al.

We apply neural nets with ReLU gates in online reinforcement learning. Our goal is to train these networks in an incremental manner, without the computationally expensive experience replay. By studying how individual neural nodes behave in online training, we recognize that the global nature of ReLU gates can cause undesirable learning interference in each node's learning behavior. We propose reducing such interferences with two efficient input transformation methods that are geometric in nature and match well the geometric property of ReLU gates. The first one is tile coding, a classic binary encoding scheme originally designed for local generalization based on the topological structure of the input space. The second one (EmECS) is a new method we introduce; it is based on geometric properties of convex sets and topological embedding of the input space into the boundary of a convex set. We discuss the behavior of the network when it operates on the transformed inputs. We also compare it experimentally with some neural nets that do not use the same input transformations, and with the classic algorithm of tile coding plus a linear function approximator, and on several online reinforcement learning tasks, we show that the neural net with tile coding or EmECS can achieve not only faster learning but also more accurate approximations. Our results strongly suggest that geometric input transformation of this type can be effective for interference reduction and takes us a step closer to fully incremental reinforcement learning with neural nets.

Sina Ghiassian, Banafsheh Rafiee, Richard S. Sutton

In this paper we present the first empirical study of the emphatic temporal-difference learning algorithm (ETD), comparing it with conventional temporal-difference learning, in particular, with linear TD(0), on on-policy and off-policy variations of the Mountain Car problem. The initial motivation for developing ETD was that it has good convergence properties under off-policy training (Sutton, Mahmood and White 2016), but it is also a new algorithm for the on-policy case. In both our on-policy and off-policy experiments, we found that each method converged to a characteristic asymptotic level of error, with ETD better than TD(0). TD(0) achieved a still lower error level temporarily before falling back to its higher asymptote, whereas ETD never showed this kind of "bounce". In the off-policy case (in which TD(0) is not guaranteed to converge), ETD was significantly slower.