Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper audits the k-NAF budget accounting mechanism in Anchored Decoding, a method for controlling KL divergence in text generation. Using both fixed and adaptive prompt search strategies, the authors evaluate whether the proxy KL divergence spend accurately reflects the true KL divergence. Results indicate that the proxy spend remains far below the allocated budget, even with adaptive search, suggesting the proxy may be overly conservative and not indicative of actual budget exhaustion.
Anchored Decoding's KL divergence proxy appears overly conservative, leaving significant budget unused even under adaptive prompt attacks.
We empirically audit the k-NAF budget-accounting mechanism in Anchored Decoding using (i) a fixed, class-stratified workload (approximately 8,500 randomized executions across six prompt classes) and (ii) an adaptive prompt-search procedure targeting high proxy spend ratios. On the fixed workload, mean cumulative KL spend remains far below the sequence-level budgets K in {600, 1000}, and an empirical Bernstein-style proxy stays below K for every class; surface-overlap diagnostics (ROUGE-L and 5-gram Jaccard) are correspondingly small. Adaptive search increases the proxy spend ratio but does not produce clear budget exhaustion. On a held-out copyright-domain workload at k = 3, several prompts exhibit proxy ratios above 1 under early-stopped evaluations with small realized sample sizes; re-evaluating the same prompts with larger allocation reduces the proxy ratio to the range [0.26, 0.40] under comparable mean spend, consistent with proxy artifacts rather than per-trajectory budget failures.