META-Learning State-based Eligibility Traces for More Sample-Efficient\n Policy Evaluation
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
Temporal-Difference (TD) learning is a standard and very successful\nreinforcement learning approach, at the core of both algorithms that learn the\nvalue of a given policy, as well as algorithms which learn how to improve\npolicies. TD-learning with eligibility traces provides a way to boost sample\nefficiency by temporal credit assignment, i.e. deciding which portion of a\nreward should be assigned to predecessor states that occurred at different\nprevious times, controlled by a parameter $\\lambda$. However, tuning this\nparameter can be time-consuming, and not tuning it can lead to inefficient\nlearning. For better sample efficiency of TD-learning, we propose a\nmeta-learning method for adjusting the eligibility trace parameter, in a\nstate-dependent manner. The adaptation is achieved with the help of auxiliary\nlearners that learn distributional information about the update targets online,\nincurring roughly the same computational complexity per step as the usual value\nlearner. Our approach can be used both in on-policy and off-policy learning. We\nprove that, under some assumptions, the proposed method improves the overall\nquality of the update targets, by minimizing the overall target error. This\nmethod can be viewed as a plugin to assist prediction with function\napproximation by meta-learning feature (observation)-based $\\lambda$ online, or\neven in the control case to assist policy improvement. Our empirical evaluation\ndemonstrates significant performance improvements, as well as improved\nrobustness of the proposed algorithm to learning rate variation.\n
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 | 0.006 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.002 |
| Bibliometrics | 0.001 | 0.002 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.004 | 0.002 |
| Research integrity | 0.001 | 0.002 |
| 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