Genetic Encoding and Shared Knowledge in Reinforcement Learning with Structured Memory
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
In partially-observable environments, agents must rely on memory, as current input alone is insufficient for decision-making. We address this challenge using Tangled Program Graphs, a genetic programming framework that supports various learning paradigms, including evolutionary reinforcement learning, where agents are structured as interconnected teams of programs for decision-making. This paper introduces a team-specific shared memory for Tangled Program Graphs that is retained throughout a reinforcement learning episode, enabling programs within the same team to exchange information during execution. This design improves coordination without interference from less-closely related programs. Additionally, each team’s shared memory and its program’s register memories are initialized with constants that are evolved through the training process, providing useful starting points to improve learning. These strategies are evaluated on MuJoCo continuous control tasks under partial observability. Results show that evolved constants and team-specific shared memory improve the fitness scores in complex tasks and perform competitively in simpler tasks. Moreover, evolved constants have notable gains when memory is not retained between timesteps. These findings highlight the importance of learned memory structures and genetic encoding in supporting adaptive behaviour in evolutionary reinforcement learning systems.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it