method
active
method:data-injection-frameworkData injection framework
Inject special task instances into pretraining data at controlled frequency.
Neighborhood — ranked by edge-count
Papers (1)
paper
Datasets (1)
dataset
- Injected 3-token comparison task.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- The view of AI as a question-answer system optimized for correctness, often inherited from supervised learning.
- Architectural choice where the original input is projected and re-injected at each recurrence, studied for its effect on fixed-point convergence
- 1984 Ashton-Tate integrated system with frames, FRED language, and overlapping windows; design reference for Playground's approach.
- Umbrella concept for the paper's dual experimental paradigms for reliably eliciting strategic deception in LLMs
- The view that pretrained LLMs are best understood as simulators capable of generating context-appropriate characters rather than single agents
- Bostrom's category of AIs that produce desired results given commands but do not act autonomously.