method
active
method:naive-samplingNaive Sampling
Baseline using Binomial(k,p) draws from precomputed rollouts.
Neighborhood — ranked by edge-count
Papers (1)
paper
- Predicting Rare LLM Failures with 30× Fewer Rolloutscitesmentions
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.
- A technique to filter model outputs; Redwood Research's project mentioned.
- The mechanism by which LLMs generate text: drawing a token from the next-token distribution and appending it to context repeatedly
- Counts c_t ~ Multinomial(S,o_t).
- A Bayesian exploration strategy that samples from the posterior distribution over model parameters to decide actions.
- Test-time compute scaling guided by reward probes with majority-vote over action.
- The paper's proposed training-free prompting strategy that prompts the model to verbalize a probability distribution over a set of responses rather than generating a single response
- AI assumption that its goals and beliefs are essentially and always true; contrasted with wise world model in Figure 3
- Prior state-of-the-art baseline using the abliterated model as proposal distribution.