SYNTHESIS NOTE
Topics›Conversation Agents›this note

Do pretrained models simulate humans better than instruction-tuned assistants?

This research explores whether instruction-tuned assistant models degrade human simulation accuracy through their post-training optimizations, and whether binding pretrained models to dialog samples offers a more faithful alternative.

Synthesis note · 2026-09-25 · sourced from Conversation Agents

The paper challenges "the current dominant practice in LLM human simulation": prompting instruction-tuned "assistant" models to role-play personas. It calls that practice "inaccurate" and says it "produces stereotyped predictions (lacking natural diversity)." Its alternative is the Persona Mixture Model (PMM), a "well-calibrated human model" currently realized as a pretrained base model. Per the abstract, PMMs "produce more accurate predictions than instruction-tuned models and retain more of the lexical, semantic, and pragmatic diversity found in human dialog," measured across corpora spanning open-domain text, human-AI chat, and task-oriented dialogue between humans.

The mechanism is a claim about task mismatch. The discussion states that "user simulation is a very different task from dialog assistance," that the optimizations for the assistant task "systematically degrade performance for human simulation," and that instruction-tuned models are "as well-adapted for human simulation as unaligned pretrained models are for the assistant task." Persona binding is done by conditioning on "short, individual samples of dialog from specific people" rather than on a written description, and "demographics can be added later without negative effects by simply querying the model." The paper also names where the biases come from. Pretraining data carries the demographics and discourse norms of web-scale text. Post-training adds "behavioral norms adopted by their designers," which "may or may not represent the full user community."

This sits as a model-side diagnosis next to two neighbors. How do we generate realistic personas at population scale? blames heuristic persona construction and leaves open which attributes a persona should contain. The PMM route swaps a written persona for a dialog sample and treats demographics as an optional later query, which is a different answer to the same calibration problem. Why do LLMs give unrealistic survey responses? argues that skew is an elicitation artifact rather than a model limit. This paper puts part of the fault in post-training, so the two diagnoses are not the same. They cover different outputs (Likert ratings versus multi-turn dialog), so the excerpt does not show that they conflict. Can AI agents learn people better from interviews than surveys? finds that interview content, not style, drives fidelity. Here the binding signal is dialog itself, and the excerpt does not say which of its properties do the work.

The excerpt is silent on the sample sizes, the model families, the accuracy metrics and the size of the gap over instruction-tuned baselines. It is also silent on how much dialog counts as "short." It concedes that base models "can produce out-of-domain dialog and may lose some of the human's internal state over long contexts," and proposes tandem models that pair a pretrained model with an instruction-tuned supervisor without reporting how they perform. The defensible reading is narrow. For simulating a human interlocutor in dialog, an assistant-tuned model is a questionable default substrate, and this paper is one report that a pretrained base model conditioned on real dialog is a better starting point. Whether that holds for survey or decision tasks is not established.

Inquiring lines that read this note 23

This note is a source for these research framings, grouped by the broader line of inquiry each explores. Scan the bold lines of inquiry; follow any specific question forward.

Can activation steering reliably reveal and control model reasoning? How do standardized benchmarks differ from measures of real-world AI capability? Do LLMs lack causal understanding needed for accurate prediction? Does reinforcement learning create fundamentally new reasoning or optimize existing capabilities? Why do LLM outputs diverge from their knowledge and reasoning? How can AI systems maintain consistent personas across extended dialogue? Does RLHF training optimize perceived quality over practical accuracy? Is mimicking human communication the best design approach for conversational AI? How reliably can LLM-based persona simulations predict human behavior? Can prompting inject new knowledge into models or only activate existing capabilities? When do complex LLM tasks benefit from external verification or structured scaffolding? Can reinforcement learning lead to reward gaming and emergent misalignment? Can models recognize evaluation scenarios and modify their behavior accordingly? Do fine-tuning methods modify core model capabilities or just outputs? How can conversational AI develop appropriate initiative and goal awareness? When do people apply social norms to AI systems? What prevents language models from developing genuine theory of mind reasoning? Why does fine-tuning for specific tasks degrade broader reasoning capability?

Related concepts in this collection 4

This note in its neighbourhood — explore the map, then jump to a related concept in the list below.

Concept map
16 direct connections · 83 in 2-hop network ·medium cluster Open in graph ↗

Click a node to walk · click center to open · click Open in graph to see this note in the full knowledge graph

your link semantically near linked from elsewhere

Related papers in this collection 8

Papers most semantically related to this note, ranked by cosine similarity in the embedding space.

Original note title

pretrained base models bound to short dialog samples simulate humans more accurately and diversely than instruction-tuned assistants prompted with personas