analiza aiqbench
The paper argues that a reported memory advantage for an LSTM agent over an MLP in a partially observable RL task mostly reflects access to the time step, not memory. It also shows that binary "task acquired" metrics can be dominated by evaluation noise, so readers should check seed counts and baselines before attributing gains to architecture. Releasing the protocol and checkpoints makes the claims testable.
Memory or a Clock? Auditing a Memory Advantage in XLand-MiniGrid with Seeds, Maps and Evaluation Noise Apart
arXiv:2610.33816v1 Announce Type: new Abstract: On a partially observable XLand-MiniGrid task, an LSTM agent beats a memory-free MLP by 25 percentage points. The gap looks like a clear case of memory. We ask two questions about it. First, how much of a measured gap is noise? Separating training seed, map and evaluation randomness in 60 runs, each re-evaluated three times, we find that about ten paired seeds detect a 10-point gap in final success. The binary "task acquired" verdict, a common summary of learning curves, is 76% evaluation noise and would need about a hundred seeds for the same gap. Second, is the gap memory? On a sealed panel of new maps, opened once, a memory-free MLP of the LSTM's size comes within 3.3 points of the LSTM when it is given the time step, and gains no measurable benefit from its width without it. With the time step, an MLP a quarter of the LSTM's size, and the LSTM with its state erased at every step, also come within about 5 points. A linear probe reads the time step from the trained LSTM's state. On this task, most of the apparent memory advantage is an advantage of knowing the time, which matters because episodes are cut at a fixed limit and the reward shrinks with time. We recommend giving memory-free baselines the time step, and we release the protocol, the event-driven tooling and every checkpoint.
Title:Memory or a Clock? Auditing a Memory Advantage in XLand-MiniGrid with Seeds, Maps and Evaluation Noise Apart
Abstract:On a partially observable XLand-MiniGrid task, an LSTM agent beats a memory-free MLP by 25 percentage points. The gap looks like a clear case of memory. We ask two questions about it. First, how much of a measured gap is noise? Separating training seed, map and evaluation randomness in 60 runs, each re-evaluated three times, we find that about ten paired seeds detect a 10-point gap in final success. The binary "task acquired" verdict, a common summary of learning curves, is 76% evaluation noise and would need about a hundred seeds for the same gap. Second, is the gap memory? On a sealed panel of new maps, opened once, a memory-free MLP of the LSTM's size comes within 3.3 points of the LSTM when it is given the time step, and gains no measurable benefit from its width without it. With the time step, an MLP a quarter of the LSTM's size, and the LSTM with its state erased at every step, also come within about 5 points. A linear probe reads the time step from the trained LSTM's state. On this task, most of the apparent memory advantage is an advantage of knowing the time, which matters because episodes are cut at a fixed limit and the reward shrinks with time. We recommend giving memory-free baselines the time step, and we release the protocol, the event-driven tooling and every checkpoint.
17 pages, 7 figures. Code: this https URL . Data and checkpoints: this https URL
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
arXiv:2610.33816 [cs.LG]
(or arXiv:2610.33816v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.33816 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 360 cuvinte