Asymmetric Capacity Allocation in Self-Refinement Pipelines

Zhuoyi Yang; Ian G. Harris; Salar Hashemitaheri; Cassie Huang; Yuangang Li; Hyunwoo Oh; Paul Dourish; Tony Givargis

cs.LGAug 21, 2026

Asymmetric Capacity Allocation in Self-Refinement Pipelines

Authors: Zhuoyi Yang, Ian G. Harris, Salar Hashemitaheri, Cassie Huang, Yuangang Li, Hyunwoo Oh, Paul Dourish, Tony Givargis,

Organizations: University of California, Irvine · Drexel University

Abstract

Self-refinement, typically structured as generation, critique, and revision, is a widely adopted paradigm for improving LLM generation and serves as a core mechanism in many LLM agents. While the three stages involve different cognitive demands, most existing approaches conveniently treat the model size as an implementation detail rather than a subject of study, which may lead to a waste of resources. Little work has systematically examined how model size affects each stage or whether effective self-refinement requires equally capable models for generation, critique, and revision. We present the first stage-wise model size study of the self-refinement pipeline on 5 benchmarks from different domains using 6 model sizes of Qwen3 and 4 model sizes of Gemma 3. We conclude that larger generators and refiners generally improve the pipeline, whereas an undersized refiner can even harm performance. Second, performance is highly insensitive to the size of the critic, although including even a small critic consistently outperforms omitting critique altogether. Our findings demonstrate that model capacity should not be allocated uniformly across self-refinement pipelines. Instead, different stages exhibit distinct size scaling characteristics, providing practical guidance for designing more computationally efficient multi-stage language model systems.

Figures & tables

Figure 1: Overview of our three-stage self-refinement pipeline with a simplified example. Given an input

x

, the generator produces an initial solution

y_{0}

, the critic provides natural-language feedback

f

x

and

y_{0}

, and the refiner revises the solution based on

x

y_{0}

, and

f

Figure 2: Stage-wise model size analysis. The three panels illustrate the generator, critic, and refiner sweeps. The highlighted stage is the stage under evaluation, whose model size is varied while the other two stages are held fixed. We also compare against a no-critique baseline.

Figure 3: Representative examples from the five benchmarks. The first column briefly illustrates the objective and input for each task, followed by the initial generation, the generated critique, and the final refined solution, demonstrating how the self-refinement pipeline operates across diverse domains.

Figure 4: Stage-wise model-size sweeps across five benchmarks under the Qwen3-32B configuration. X-axis is the model sizes. Blue line, green line, and the red line represent Generator sweep, Critic sweep and the Refiner sweep. The grey dashed line reports the corresponding initial-generation performance without refinement. Steeper curves indicate greater sensitivity to the model size of that stage. Metrics are defined in Section 4.2 .

Benchmark	Model	Generator		Critic		Refiner
Benchmark	Model	$\Delta$	Std.	$\Delta$	Std.	$\Delta$	Std.
Meeting Planning	Qwen	16.00	6.15	9.00	3.10	32.00	11.20
Meeting Planning	Gemma	24.00	8.76	8.00	2.92	34.00	12.64
CNN/DailyMail	Qwen	10.00	3.95	8.00	3.10	23.40	8.99
CNN/DailyMail	Gemma	25.10	9.62	6.30	2.24	16.80	6.39
ZebraLogic	Qwen	8.65	3.07	3.89	1.46	13.39	4.71

Table 1: Performance variability across model sizes when varying one stage while fixing the other two at the largest model. Values are reported in each benchmark’s evaluation metric, as specified in Section 4.2 .

Figure 5: Distribution of manual critique-quality scores across five benchmarks. For each benchmark (top to bottom: Meeting Planning, CNN/DailyMail, ZebraLogic, PIE, and CollaboSentGen): Stacked segments show the percentage of critiques assigned scores 1–5, while the percentages on the right report the fraction of non-misleading critiques (scores 3–5).

Input	Start: The Castro (9:00 AM) Ronald: Chinatown (9:30–11:30, 75 min) Nancy: Nob Hill (11:15–12:00, 30 min)
Generated	09:30–10:30 Ronald 10:50–11:20 Nancy
0.6B Critic (Score 4)	Detects: Ronald’s meeting is only 60 min (requires 75 min).
32B Critic (Score 5)	Detects: Ronald’s meeting is only 60 min. Detects: Nancy is scheduled outside her availability window.

Table 2: Representative example illustrating performance of 0.6B critic and 32B critic.

Benchmark	Generator Sweep	Refiner Sweep
Meeting Planning	+9.67	+10.34
CNN/DailyMail	+4.56	+5.21
ZebraLogic	+7.97	+7.86
PIE	+9.03	+11.06
CollaboSentGen	+3.71	+4.01

Table 3: Average critique gain with Qwen3-0.6B as the critic over the matched no-critique baseline across the generator and refiner sweeps.

Figure 6: A degradation example from ZebraLogic. Despite a correct critique, the 0.6B refiner is introducing new errors while the 32B refiner is incorporating the critique.

Appendix figures & tables19 assets

Supplementary material from the paper’s appendix.

Appendix

Figure 7: Stage-wise model-size sweeps across five benchmarks under the Gemma 3-27 configuration. X-axis is the model sizes. Blue line, green line, and the red line represent Generator sweep, Critic sweep and the Refiner sweep respectively. The grey dashed line reports the corresponding initial-generation performance without refinement. Steeper curves indicate greater sensitivity to the model size of that stage. Metrics are defined in Section 4.2

Figure 8: Stage-wise model-size sweeps across five benchmarks under the Qwen3-14B configuration. X-axis is the model sizes. Blue line, green line, and the red line represent Generator sweep, Critic sweep and the Refiner sweep respectively. The grey dashed line reports the corresponding initial-generation performance without refinement. Steeper curves indicate greater sensitivity to the model size of that stage. Metrics are defined in Section 4.2

Figure 9: Stage-wise model-size sweeps across five benchmarks under the Qwen3-8B configuration. X-axis is the model sizes. Blue line, green line, and the red line represent Generator sweep, Critic sweep and the Refiner sweep respectively. The grey dashed line reports the corresponding initial-generation performance without refinement. Steeper curves indicate greater sensitivity to the model size of that stage. Metrics are defined in Section 4.2

Table 4: Refiner sizes that perform below the corresponding initial generation under different model configurations. A checkmark indicates that using the corresponding refiner size results in lower performance than the initial generation.

Figure 10: Generator prompt for Meeting Plan

Figure 11: Critic prompt for Meeting Plan

Figure 12: Refiner prompt for Meeting Plan

Figure 13: Generator prompt for cnn/DailyMail

Figure 14: Critic prompt for cnn/DailyMail

Figure 15: Refiner prompt for cnn/DailyMail

Figure 16: Generator prompt for ZebraLogic

Figure 17: Critic prompt for ZebraLogic

Figure 18: Refiner prompt for ZebraLogic

Figure 19: Generator prompt for PIE

Figure 20: Critic prompt for PIE

Figure 21: Refiner prompt for PIE

Figure 22: Generator prompt for CollaboSentGen

Figure 23: Critic prompt for CollaboSentGen

Figure 24: Refiner prompt for CollaboSentGen

LLM Self-Refinement

Large language model self-refinement studies how models can improve outputs through repeated critique, verification, revision, and feedback.

See all 115 papers

Streamlined Reflective Evolution for Task-Adaptive Self-Refinement Pipelines
Sep 26, 2026Xiaofan Zhou, Lu ChengLLM Self-Refinement LLM Agent Workflow Optimization
Pennsylvania State University
Evaluating and Improving LLM Self-Modeling
Aug 31, 2026Siqi Zeng, Andre N. Assis, Rowan WangLLM Evaluation Language Model Introspection
University of Illinois · Constellation · Anthropic
Compute Has Shape: Self-Morphing the Internal Computation of Frozen LLMs
Oct 28, 2026Junxi Wu, Yiwen Guo, Shu-Tao XiaInference-Time Adaptation of Language Models Test-Time Scaling
School of Statistics and Data Science, Nankai University · Tsinghua Shenzhen International Graduate School, Tsinghua University · Independent Researcher

Sep 26, 2026J/K move · Enter open · S save

Streamlined Reflective Evolution for Task-Adaptive Self-Refinement Pipelines

Xiaofan Zhou, Lu Cheng

Pennsylvania State University

LLM Self-Refinement LLM Agent Workflow Optimization Automatic Prompt Optimization

Open paper AskPDF arXiv

Reflective prompt optimization improves large language model (LLM) systems without updating model weights, but fixed architectures constrain how self-refinement is organized. We introduce Workflow-Designing Agents (WDA), a framework for streamlined reflective evolution of task-adaptive self-refinement pipelines. Starting from a minimal prompt, WDA jointly evolves stage instructions and their sequential structure. During evolution, we find that repeated revisions can accumulate redundant instructions in a single prompt. In WDA, we propose to address this problem with SPLIT, which redistributes these instructions across specialized stages. Three-example reflection and local screening guide selective search, while calibration scores guide Pareto admission and rollback of unhelpful trailing updates. The resulting pipelines are task-adaptive: their instructions and depth are learned from task data, then fixed for all test inputs within that task. We evaluate WDA on five benchmarks spanning knowledge, mathematical reasoning, multi-hop question answering, and instruction following. On Qwen3.5-9B, WDA achieves an average score of 51.24%, improving over the initial solver by 8.63 percentage points and the variant without SPLIT by 2.60 points. On GPT-4.1-mini, it achieves 49.00%, with corresponding gains of 5.62 and 3.69 points. These results support task-adaptive self-refinement as a complementary direction to broader agentic workflow search.

Abstract

LLM Self-Refinement

Explore similar work

Streamlined Reflective Evolution for Task-Adaptive Self-Refinement Pipelines

Evaluating and Improving LLM Self-Modeling

Compute Has Shape: Self-Morphing the Internal Computation of Frozen LLMs