# Elicitation without Backpropagation: Steering Model Behavior by Optimizing the Latent Posterior

arXiv: https://arxiv.org/abs/2607.18804 · HTML: https://mlsanity.com/paper/2607.18804/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 25 of its 28 references that are in the map.
- First posted on arXiv 21 Jul 2026.

## Builds on (most cited first, 25 of 25)

- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Prefix-Tuning: Optimizing Continuous Prompts for Generation](https://mlsanity.com/paper/2101.00190/) (2021)
- [The Power of Scale for Parameter-Efficient Prompt Tuning](https://mlsanity.com/paper/2104.08691/) (2021)
- [An Explanation of In-context Learning as Implicit Bayesian Inference](https://mlsanity.com/paper/2111.02080/) (2021)
- [What Can Transformers Learn In-Context? A Case Study of Simple Function Classes](https://mlsanity.com/paper/2208.01066/) (2022)
- [What learning algorithm is in-context learning? Investigations with linear models](https://mlsanity.com/paper/2211.15661/) (2022)
- [Toxicity in ChatGPT: Analyzing Persona-assigned Language Models](https://mlsanity.com/paper/2304.05335/) (2023)
- [Scalable and Transferable Black-Box Jailbreaks for Language Models via Persona Modulation](https://mlsanity.com/paper/2311.03348/) (2023)
- [Persona Features Control Emergent Misalignment](https://mlsanity.com/paper/2506.19823/) (2025)
- [In-Context Learning through the Bayesian Prism](https://mlsanity.com/paper/2306.04891/) (2023)
- [Transformers represent belief state geometry in their residual stream](https://mlsanity.com/paper/2405.15943/) (2024)
- [Meta-learning of Sequential Strategies](https://mlsanity.com/paper/1905.03030/) (2019)
- [Personas as a Way to Model Truthfulness in Language Models](https://mlsanity.com/paper/2310.18168/) (2023)
- [Martingale posterior distributions](https://mlsanity.com/paper/2103.15671/) (2021)
- [Who's asking? User personas and the mechanics of latent misalignment](https://mlsanity.com/paper/2406.12094/) (2024)
- [Meta-trained agents implement Bayes-optimal agents](https://mlsanity.com/paper/2010.11223/) (2020)
- [Learning Universal Predictors](https://mlsanity.com/paper/2401.14953/) (2024)
- [Memory-Based Meta-Learning on Non-Stationary Distributions](https://mlsanity.com/paper/2302.03067/) (2023)
- [Constrained belief updates explain geometric structures in transformer representations](https://mlsanity.com/paper/2502.01954/) (2025)
- [TabMGP: Martingale Posterior with TabPFN](https://mlsanity.com/paper/2510.25154/) (2025)
- [Understanding Prompt Tuning and In-Context Learning via Meta-Learning](https://mlsanity.com/paper/2505.17010/) (2025)
- [Uncertainty Decomposition for Bayes-Filtered Transformers via Bayesian Predictive Inference](https://mlsanity.com/paper/2602.04596/) (2026)
- [Next-token pretraining implies in-context learning](https://mlsanity.com/paper/2505.18373/) (2025)
- [Why is prompting hard? Understanding prompts on binary sequence predictors](https://mlsanity.com/paper/2502.10760/) (2025)
- [What does a Bayes-filtered transformer believe? A predictive Monte Carlo approach](https://mlsanity.com/paper/2607.17060/) (2026)

## Abstract (arXiv)

In the latent posterior model of transformer behavior, the next-token distribution arises from a posterior over latent predictive models conditioned on the context, mixed to generate continuations. We exploit this model in settings where it is exact, namely Bayes-filtered transformers (BFTs) meta-learned on sequences from a hierarchical prior, to introduce Posterior Prefix Tuning (PPT), a new method for eliciting behavior from a transformer: given a utility function on continuations, find a prompt under which the transformer generates continuations of high expected utility. For a BFT, the elicitation objective factors through the latent posterior, and the gradient of this objective can be estimated from samples of the prior alone. PPT optimizes the parameters of a distribution over hard prompts: it draws prior samples once from the BFT via predictive Monte Carlo (PMC), then estimates the gradient by importance sampling against them. The optimization performs no transformer forward passes and no backpropagation through the transformer, and the prior samples are utility-independent, so a single set of samples drives elicitation against any number of utilities at negligible marginal cost. We validate PPT on Beta--Bernoulli and reinforced urn BFTs across three utility families (reverse cross-entropy, frequency matching, Dyck validity).