A novel cognitively-inspired hybrid model is presented which optimally trades off simulation and reasoning to improve model performance and computational efficiency and draws a parallel with this to resource-rational theories of human cognition, which propose that people optimally trade off simulation and reasoning dep...
Ying-Qiao Wang, Eric J. Bigelow, Tomer D. Ullman et al.· 0 citations
Recursive self-improvement (RSI) of a model on non-verifiable tasks, such as open-ended research, faces a supervision bottleneck when its outputs exceed what even human experts can reliably assess, leaving the model itself (optimizee) as the best available optimizer and evaluator. However, a single model instance strug...
Hyunin Lee, Jinglue Xu, Jeffrey Seely et al.· 0 citations
The proposed symbiotic architecture employs an injector module that writes audio-conditioned vectors directly into the target LLM's short-term memory, enabling the LLM to behave as an audio language model (ALM) and improves the scalability of ALMs.
SAIL is a framework that reframes robot imitation as an iterative refinement problem capable of scaling with test-time compute, and utilizes Monte Carlo Tree Search, where each node is a complete trajectory and edges correspond to trajectory refinements.
The value of execution-improved software as a resource for acquiring new policies in this setting is demonstrated, although initial references remain better on two target tasks when averaged across runs.
Inspired by dual-coding theory, this work proposes a memory architecture that uses parallel visual and verbal codes, which it calls DualMem, and views this as a step towards memory systems that preserve a richer record of agents'observations.
The findings suggest world knowledge and task-directed ability can be learned in geometrically complementary forms, and that future post-training pipelines should consider how best to engineer the interface between them.
Rui-Ze Xu, Xiao Yu, Yu-Jin Tang et al.· 0 citations
Recursive Harness Self-Improvement is introduced, which represents the harness as a prompt-level specification of the agent loop and iteratively refines it using pairwise feedback over its own revision history, suggesting RHI as a practical algorithm for continual learning within the paradigm of model--harness co-evolu...
