Inference-Time Adaptation

APO

D1Context Inj.
D2Behav. Ctrl.Task Spec.
D3Few Demo.Zero-Shot
D4Ver.-Persist.
D5I/O Space
D6LLMMLLM

In brief — site editorial, not from the paper

Automatically searches for better prompt wording, optimising the instruction rather than the model.

Definition

APO frameworks, such as DSPy [106] or OPRO [219], replace manual prompt engineering with algorithmic search, treating language models as programmable engines. APO utilizes Context Injection within the Input/Output-Space, meaning the underlying model parameters remain completely frozen. However, because the optimization compiles a tested, metric-optimized prompt pipeline, the resulting artifact is classified as Version-Persistent rather than Session-Ephemeral. Data requirements are flexible: APO can utilize Few Demonstrations to bootstrap the optimization, but it can also operate strictly Zero-Shot by relying entirely on programmatic evaluation metrics or an "LLM-as-a-judge" [115, 231] to score unannotated inputs. Ultimately, this algorithmic search aims for both Task Specialization (optimizing accuracy on a target task) and Behavior Control (systematically enforcing strict output formats, stylistic guidelines, or reasoning templates without altering weights).

Verbatim from the paper — Inference-Time Adaptation

Nearest profiles

PE (0.28), CE (0.33), RAG (0.33), ICL (0.36), Test-Time Compute Scaling (0.58)

Computed from the taxonomy data (Gower distance over all six dimensions)

References

  1. [106] Omar Khattab et al. 2024. DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelines.
  2. [115] Dawei Li et al. 2025. From Generation to Judgment: Opportunities and Challenges of LLM-as-a-judge. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing
  3. [219] Chengrun Yang et al. 2024. Large Language Models as Optimizers. arXiv
  4. [231] Lianmin Zheng et al. 2023. Judging LLM-as-a-judge with MT-bench and Chatbot Arena. Proceedings of the 37th International Conference on Neural Information Processing Systems

Numbered as in the paper

Cite this row

APO — six-dimensional profile
D1: context-injection
D2: behavior-control, task-specialization
D3: few-demonstrations, zero-shot
D4: version-persistent
D5: input-output-space
D6: llm, mllm
Source: arXiv:2608.06246
Compare with another technique Find in the explorer