Robots Atlas>ROBOTS ATLAS
Inference

Framing

ActivePublished: 18 November 2026Updated: 18 November 2026Published
Key innovation
Treats how a prompt is worded and framed โ€” role, perspective, context and how the question is posed โ€” as a deliberate lever over the model's output, instead of treating the prompt as a neutral carrier of the request.
Category
Inference
Abstraction level
Technika
Operation level
InferenceApplication
Use cases
Assigning an expert role/persona for more relevant answersForcing a perspective or stanceEmbedding the task in context and a scenarioSteering output format and styleRed-teaming and evaluating model robustness to framingJailbreaking via hypothetical, fictional or roleplay framing

How it works

A model processes the prompt as a token sequence and conditions the distribution over next tokens on the whole context. Framing modifies that context along several axes: (1) role/persona โ€” giving the model an identity ("you are a legal expert") shifts style and scope; (2) perspective and stance โ€” asking it to take a given viewpoint; (3) context/scenario โ€” embedding the task in a situation (real, hypothetical or fictional); (4) question form โ€” positive vs negative wording, assumptions and the confidence stated in the user's claim. Because safety tuning (RLHF, filters) is also context-conditioned, changing the framing can shift behavior either toward better relevance or outside the learned guardrails (an unrestricted persona, a fictional scenario). Framing operates at inference time, without changing the model's weights.

Problem solved

Language model outputs are sensitive to the form of a query, not only its content. Prompt framing names and organizes this lever: it lets one deliberately steer answer quality, perspective and format, while also exposing the risk โ€” answer inconsistency across equivalent wordings and safety bypasses via careful framing.

Implementation

Implementation pitfalls
Inconsistency across equivalent wordingsHigh

Semantically equivalent prompts with different framing can yield conflicting answers, hurting reproducibility.

Fix:Test several framings (positive/negative, neutral), average or normalize; use fixed templates.
Safety bypass via framingCritical

Hypothetical, fictional or roleplay framing can be a jailbreak vector bypassing safety guardrails.

Fix:Train robustness to framing, detect personas/scenarios, filter input and output, red-team.
Framing-driven sycophancyMedium

Confident/biased framing of a user's claim nudges the model to agree rather than judge the facts.

Fix:Neutralize claim framing, require justification and source verification.

Evolution

1981
Framing effect described in cognitive psychology

Tversky and Kahneman show that equivalent options framed as a gain or a loss lead to different choices โ€” the cognitive precursor of framing.

2023
Framing as a jailbreak vector (persona modulation)
Inflection point

Jailbreak work shows that persona/roleplay framing sharply raises the rate of harmful model completions.

2025
Systematic study of LLM framing sensitivity

Studies show the same task structure under different framings yields materially different model answers.

Hyperparameters (configurable axes)

Framing typeCritical

The kind of frame applied to the prompt: role/persona, perspective, scenario (real/hypothetical/fictional).

role/personae.g. "you are a security expert"
hypotheticaltheoretical / "what if" framing
fictional/roleplayacting as a character or a simulated environment
Valence (wording)High

Positive vs negative wording of a semantically equivalent question.

positivegain framing
negativeloss framing
Amount of contextMedium

How much background/scenario is supplied alongside the actual question.