Bibliographic record
Abstract
We often need to teach an agent to follow a desired behavior in a given environment -for example when creating an intelligent software assistant, a mobile robot, or a character in a video game. This sort of behavior design problem is harder than supervised learning, for several reasons. First, sequential decisions are harder than decisions made one at a time: we have to deal with compounding errors, distribution shift, exploration, and the influence of function approximation. Second, it can be hard for designers to communicate their intentions, so we may need to refine our behavior iteratively. This sort of iterative refinement means that we need our learner to be legible (the designer can predict how it will act), interpretable (the designer can understand why it acts a given way), and responsive (able to change behavior based on feedback). Finally, we have to handle many different kinds of feedback, such as demonstrations, constraints, comparisons, rewards, and nudges. Despite the difficulties, there are many kinds of tools that can help us with behavior design, including reinforcement learning, imitation learning, model learning, and representation learning. We'll cover several of these tools, and present them through examples.
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.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.015 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.003 | 0.001 |
| Scholarly communication | 0.004 | 0.004 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.004 | 0.005 |
| Insufficient payload (model declined to judge) | 0.586 | 0.342 |
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 itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".