Pith. sign in

REVIEW 3 major objections 4 minor 38 references

FLARE: Few-shot Learning-based Adaptive Reflective Engine

T0 review · 3 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read FLARE beats the prior optimizer on every task-model pair it tests, with gains up to +15.3 points.

desk verdict FLARE's error-aware prompt optimization is a real idea with careful experiments, but test-based configuration selection and memorized validation examples undermine the headline gains. read the letter →

arxiv 2608.02919 v1 pith:4AK6KVBV submitted 2026-08-03 cs.CL

Dhanasekar Sundararaman , Bharat Gandhi , Aashna Garg , Minjie Li This is my paper · ORCID
classification cs.CL
keywords promptoptimizationfew-shotlearningreflectivefeedbackerror-awarelargelanguagemodelsretrieval-augmentedgenerationtoolcallingmulti-labelemotionclassification
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper sets out to show that automated prompt optimization works best when the optimizer is grounded in concrete, per-example failures rather than in abstract instructions alone. The proposed method, FLARE, repeatedly diagnoses why a model got each validation example wrong and rewrites the prompt to fix those specific error patterns, helped by a small set of few-shot reference examples. Against a genetic-Pareto reflective-evolution baseline, the authors report winning on all ten task-model combinations they evaluate: up to +14.2 points on HotPotQA, 87.0% versus 81.0% on tool calling, and +15.3 micro-F1 on GoEmotions. They also report that FLARE reaches its best emotion-classification score with only 100 validation examples and is more stable across random seeds. The claim, if correct, is that reflective instruction evolution without few-shot grounding is leaving large, accessible gains on the table.

What carries the argument

The load-bearing mechanism is the error-aware reflective feedback loop. For each validation example the optimizer receives a signal CORRECT or WRONG($y_i$, $\hat{y}_i$) with the true label and the model's wrong prediction, and a strong LLM meta-optimizer performs a thinking pass to diagnose the root cause and rewrite the prompt surgically. The loop is stabilized by a sliding window of the last three iterations' scores and prompt summaries, and by few-shot reference examples held fixed. For tool calling, multi-tool ground-truth sequences are decomposed into individual tool-call examples so the optimizer can tell whether failure is in tool selection, argument extraction, or formatting. This machinery is what grounds abstract reflection in concrete failures.

What would settle it

Take the final optimized prompt documented in the appendix and run it on a held-out set that excludes or paraphrases the exact strings the prompt memorizes, such as 'BLOCK BLOCK BLOCK him!' and 'downvoted heavily but i understand your experience. salute, soldier.'. If the GoEmotions gain collapses toward the baseline on those paraphrases, the advantage is memorization; if the gain persists, it is a generalizable rule.

Watch

Extended reading notes

Core claim

The central claim, stated on the paper's own terms, is that reflective prompt optimization should be driven by explicit per-instance error signals. At each iteration, a frontier model acting as meta-optimizer looks at the current best prompt, the complete training set, the model's prediction and the ground-truth label for every validation example, the current validation score, and a sliding history of recent prompt-score pairs, then writes a corrected prompt that directly targets the diagnosed failure modes. The paper claims this error-aware, few-shot-grounded loop consistently outperforms an instruction-only Pareto-evolution baseline on every task-model pair tested, across retrieval-augmented reasoning, tool calling, and multi-label emotion classification, and that the advantages come from the quality of the search rather than from scale: a heavier baseline budget does not close the gap, and peak performance is reached with as little as 100 validation examples.

Load-bearing premise

The reported gains are assumed to come from improved, generalizable prompt rules rather than from the optimizer memorizing the specific validation examples it saw during optimization.

Editorial extensions

If this is right

  • If FLARE's results hold, instruction-only reflective prompt evolution is leaving substantial accuracy on the table across RAG, tool calling, and classification.
  • Prompt optimization can be highly data-efficient: on GoEmotions the best prompt comes from 100 validation examples, so low-label settings are a natural application target.
  • Throwing more optimization budget at an evolutionary search does not substitute for a better search signal, since the heavier-budget baseline still trails FLARE at nearly every validation size.
  • The same steep-rise-then-plateau efficiency profile appears on two different GPT-5-series backbones, so the data-efficiency property is attributed to the method rather than the model.
  • Smaller gains on multi-hop reasoning suggest that errors deep in the retrieval chain cannot be fully repaired by prompt editing alone, pointing toward hybrid prompt-plus-retriever optimization.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Inference: the appendix's special override cases show the optimizer memorizing exact validation strings, such as 'BLOCK BLOCK BLOCK him!', into the final prompt; how much of the headline gain comes from such memorization versus transferable rules is left unmeasured, and a paraphrased re-test would settle it.
  • Inference: the decomposition of multi-tool traces into per-tool examples is a general trick that could isolate failure types in other agentic pipelines, not just customer-service tool calling.
  • Inference: the error-aware loop's success with 100 examples suggests a practical recipe for quickly adapting prompts in new domains where labels are scarce, though the paper only demonstrates this on emotion classification.
  • Inference: FLARE's per-iteration full validation rescoring costs roughly 2.3 times the baseline's evaluation calls; if prompts are deployed widely, the authors' one-time offline cost framing invites a cost-benefit study of prompt optimization in long-lived production systems.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper introduces FLARE, an iterative prompt-optimization framework in which a GPT-5-series meta-optimizer rewrites a task prompt based on per-instance error signals computed over a validation set, together with a small set of few-shot reference examples. The method is compared with GEPA, Promptomatix, and OpenAI's Prompt Optimizer on multi-label emotion classification (GoEmotions), a tool-calling benchmark (tau2-bench), and three retrieval-augmented reasoning tasks (HotPotQA, MedQA, 2WikiMultiHopQA). The central claims are that FLARE consistently outperforms GEPA on every task-model pair, reaches its GoEmotions peak with as few as 100 validation examples, and is more stable across seeds than GEPA. Appendix A documents the iterative prompt-optimization process, including the final prompt that was selected as validation-best.

Significance. If the reported results are reliable, FLARE would be a meaningful contribution to prompt optimization: it challenges GEPA's claim that instruction-only reflective evolution outperforms few-shot-grounded optimization, and its validation-set-size sweep provides a useful data-efficiency analysis. The paper's strengths include running configurations over three seeds with reported standard deviations, comparing multiple baselines, including an appendix with actual iterative prompts, and releasing code. However, two protocol issues bear directly on the central generalizability and data-efficiency claims: the headline GoEmotions configuration is selected using test-set performance, and the returned prompt contains verbatim validation examples as memorized override cases. These issues make the reported gains, as they stand, an optimistic upper bound rather than a reliable estimate of FLARE's generalizable performance.

major comments (3)
  1. [Section 5.2 / Table 2 / Algorithm 1] The headline GoEmotions number is selected from a test-based validation-size sweep. Algorithm 1 states that the test set is 'tracked for reporting only,' but Table 1 reports FLARE's best configuration across the validation-set-size sweep, and Table 2 evaluates all configurations on the full held-out test split. Selecting the validation-set size (and GEPA's light/heavy budget) from test scores means the reported +15.3 gain in Table 1 is a maximum over eight validation sizes and multiple seeds, not an unbiased estimate of FLARE's expected performance. The authors should re-select the configuration using validation-only scores and report the resulting test performance; the current protocol cannot support the claim that FLARE 'reaches its peak performance using as few as 100 validation examples.'
  2. [Appendix A, Iteration 20] The final prompt selected as validation-best contains a section titled 'SPECIAL OVERRIDE CASES (MEMORIZE)' that hard-codes exact validation texts such as 'BLOCK BLOCK BLOCK him!' and 'downvoted heavily but i understand your experience. salute, soldier.' with forced labels. This is direct evidence that the optimizer is memorizing validation instances rather than learning generalizable instruction rules. Even if the test set is disjoint, this overfitting to the validation distribution can inflate test scores when the test distribution is similar, and it directly undermines the paper's data-efficiency interpretation. The authors should quantify performance when these memorized override cases are removed, or adopt a protocol that explicitly prohibits verbatim validation examples from appearing in the returned prompt.
  3. [Section 5.2 / Figure 1] The claim that FLARE is 'markedly more stable across seeds than GEPA' is based on comparing standard deviations of the selected best configurations, but the selection process itself uses test scores. Since GEPA's light/heavy budgets and the validation size are selected after seeing test results, the reported stability comparison is also affected by selection effects. A fairer comparison would hold the selection protocol fixed for both methods and report the stability of the selected configuration under that protocol.
minor comments (4)
  1. [Appendix A, Iteration 1] There is a typo in the heading 'CURIOUSITY [7] vs NEUTRAL [27]'; it should be 'CURIOSITY.'
  2. [Section 4.2] The RAG composite metric (0.3 x retrieval + 0.7 x generation quality) is not fully defined; please specify the exact retrieval and generation submetrics and how they are normalized before combination.
  3. [Section 3, Algorithm 1] Algorithm 1 passes the complete training set T to the meta-optimizer but does not specify how many reference exemplars are sampled or how they are selected; Section 4.1 mentions a fixed reference exemplar set of 50, but the relationship between T and the few-shot reference examples should be clarified in the algorithm pseudocode.
  4. [Table 4 caption] The caption for Table 4 says 'Best optimized scores per dataset-LLM combination' but the RAG comparison is restricted to FLARE and GEPA; the caption should state this restriction explicitly, as is done in the body text.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: FLARE is an empirical optimizer comparison grounded in external datasets and independent baselines.

full rationale

FLARE is an empirical prompt-optimization paper rather than a derivation, so the circularity patterns enumerated here do not apply. The algorithm's inputs are external training and validation examples plus a task metric; its output is a prompt; and all headline numbers are held-out test measurements taken after validation-based selection. No equation defines the reported test score in terms of the validation score, and no fitted parameter is relabeled as a prediction. The comparison against GEPA rests on independent published work by other authors, and every cited framework (DSPy, GEPA, Promptomatix, OpenAI Prompt Optimizer) is an external baseline rather than a self-citation chain. The two genuine protocol concerns—Appendix A's verbatim 'SPECIAL OVERRIDE CASES (MEMORIZE)' examples and the Table 1 practice of reporting each method's best configuration across the validation-set-size sweep—are evidence of possible overfitting and optimistic test-based selection, which threaten generalizability and statistical validity. However, overfitting and selection bias are not circular reasoning: the reported test scores remain measurements produced by holding out a disjoint test split, not consequences entailed by construction from the optimization inputs. Per the hard rules, correctness-risk concerns such as 'this is not standard consensus' or 'this may be overfit' do not raise the circularity score. The derivation chain—optimize on validation, measure on test, compare against external baselines—is self-contained and non-circular.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claim rests on an empirical comparison, so the ledger mainly records assumptions about test-set hygiene, metric validity, baseline fidelity, and stochastic evaluation. The most consequential item is the first one: the paper's own procedure violates the stated test-set-isolation assumption when it selects the best validation size from test scores.

free parameters (4)
  • Validation-set size (best configuration) = 100 for GPT-5.1, 1000 for GPT-5-Chat on GoEmotions
    The reported headline gains choose the row with the highest test micro-F1 in Table 2, so the hyperparameter is fit to the test set.
  • Optimization iterations N = 40
    Fixed for all tasks; no ablation shows sensitivity.
  • Reference exemplar set size = 50
    Fixed at 50 examples; no ablation of this number is reported.
  • Decoding temperature = 1.0
    Used for both optimizer and evaluator to encourage exploration; chosen by hand.
assumptions (5)
  • domain assumption The held-out test set is never used for configuration selection.
    The paper asserts this in Section 3, but Table 1 selects the best validation-set size from Table 2, which reports test micro-F1.
  • domain assumption The evaluation metric for RAG (0.3 times retrieval plus 0.7 times generation) is a valid measure of prompt quality.
    Defined in Section 4.2; no external validation of the composite metric is provided, and it differs from standard RAG accuracy metrics.
  • ad hoc to paper Overfitting to validation examples is benign because the test set is statistically similar.
    Appendix A's final prompt hard-codes exact validation texts; the paper assumes this transfers to test without measuring overlap.
  • domain assumption GEPA is implemented faithfully and the comparison budgets are fair.
    The paper relies on GEPA as the central baseline, but no code or hyperparameter settings for GEPA are included beyond light/heavy budgets.
  • domain assumption GPT-5 API outputs are stable enough that 3 seeded runs estimate true performance.
    Temperature 1.0 is used for evaluation; runs are stochastic, and no determinism or larger repetition count is used.

how reviews work

0 comments
Cite this review

Pith. "Pith review of FLARE: Few-shot Learning-based Adaptive Reflective Engine." pith.science (2026). https://pith.science/paper/4AK6KVBV

@misc{pith2026260802919,
  author       = {Pith},
  title        = {Pith review of: FLARE: Few-shot Learning-based Adaptive Reflective Engine},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4AK6KVBV}},
  note         = {Machine review of arXiv:2608.02919}
}
read the original abstract

Large language models (LLMs) are increasingly deployed in complex, compound AI systems where performance hinges on the quality of prompts. Recent state-of-the-art optimizers like GEPA (Genetic-Pareto) have argued that reflective instruction evolution can outperform traditional reinforcement learning and few-shot optimization. In this work, we challenge this shift by introducing FLARE (Few-shot Learning-based Adaptive Reflective Engine), a framework that leverages advanced reflective mechanisms and a small set of few-shot reference examples to optimize instructions. We evaluate our method across a diverse suite of benchmarks -- spanning retrieval-augmented reasoning (HotPotQA, MedQA, 2WikiMultiHopQA), tool calling, and multi-label emotion classification (GoEmotions) -- using the GPT-5 series of models. Our results demonstrate that FLARE consistently outperforms GEPA, winning on every task-model pair: it achieves gains of up to +14.2 points on HotPotQA (52.2 vs. GEPA's 42.2 with GPT-5-Chat), reaches 87.0% on tool calling (vs. 81.0% for GEPA), and lifts GoEmotions micro-F1 to 52.7% (+15.3) with GPT-5.1 on the full 5408-example test split, more than doubling GEPA's +5.7 gain. Beyond raw accuracy, FLARE is also strikingly data-efficient: on GoEmotions it reaches its peak performance using as few as 100 validation examples, while remaining markedly more stable across random seeds than GEPA. Our findings suggest that while reflective instructions are powerful, the strategic optimization of few-shot learning remains a critical frontier for maximizing the potential of next-generation LLMs.

Figures

Figures reproduced from arXiv: 2608.02919 by the authors.

Figure 1
Figure 1. Data efficiency on the GoEmotions classification task. Test micro-F1 (%) on the full held-out split (5,408 examples) as a [PITH_FULL_IMAGE:figures/full_fig_p005_1.png] view at source ↗

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

38 extracted references · 19 canonical work pages

  1. [1]

    Agrawal, Shangyin Tan, Dilara Soylu, Noah Ziems, Rishi Khare, Krista Opsahl-Ong, Arnav Singhvi, Herumb Shandilya, Michael J

    Lakshya A. Agrawal, Shangyin Tan, Dilara Soylu, Noah Ziems, Rishi Khare, Krista Opsahl-Ong, Arnav Singhvi, Herumb Shandilya, Michael J. Ryan, Meng Jiang, Christopher Potts, Koushik Sen, Alexandros G. Dimakis, Ion Stoica, Dan Klein, Matei Zaharia, and Omar Khattab. 2025. GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning. arXiv:2507.19...

    arXiv 2025

  2. [2]

    Akari Asai, Zeqiu Wu, Yizhong Wang, Avirup Sil, and Hannaneh Hajishirzi. 2023. Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection. arXiv:2310.11511 https://arxiv.org/abs/2310.11511

    arXiv 2023

  3. [3]

    2025.𝜏 2-Bench: Evaluating Conversational Agents in a Dual-Control Environ- ment

    Victor Barres, Honghua Dong, Soham Ray, Xujie Si, and Karthik Narasimhan. 2025.𝜏 2-Bench: Evaluating Conversational Agents in a Dual-Control Environ- ment. arXiv:2506.07982 https://arxiv.org/abs/2506.07982

    arXiv 2025

  4. [4]

    Dorottya Demszky, Dana Movshovitz-Attias, Jeongwoo Ko, Alan Cowen, Gaurav Nemade, and Sujith Ravi. 2020. GoEmotions: A Dataset of Fine-Grained Emotions. arXiv:2005.00547 https://arxiv.org/abs/2005.00547

    arXiv 2020

  5. [5]

    Xanh Ho, Anh-Khoa Duong Nguyen, Saku Sugawara, and Akiko Aizawa. 2020. Constructing a Multi-Hop QA Dataset for Comprehensive Evaluation of Reason- ing Steps. InProceedings of the 28th International Conference on Computational Linguistics (COLING). doi:10.18653/v1/2020.coling-main.580

    doi:10.18653/v1/2020.coling-main.580 2020

  6. [6]

    Ziyan Jiang, Xueguang Ma, and Wenhu Chen. 2024. LongRAG: Enhancing Retrieval-Augmented Generation with Long-context LLMs. arXiv:2406.15319 https://arxiv.org/abs/2406.15319

    arXiv 2024

  7. [7]

    Xu, Luyu Gao, Zhiqing Sun, Qian Liu, Jane Dwivedi-Yu, Yiming Yang, Jamie Callan, and Graham Neubig

    Zhengbao Jiang, Frank F. Xu, Luyu Gao, Zhiqing Sun, Qian Liu, Jane Dwivedi-Yu, Yiming Yang, Jamie Callan, and Graham Neubig. 2023. Active Retrieval Aug- mented Generation. InProceedings of the 2023 Conference on Empirical Methods in Natural Language Processing (EMNLP). doi:10.18653/v1/2023.emnlp-main.495

    doi:10.18653/v1/2023.emnlp-main.495 2023

  8. [8]

    Di Jin, Eileen Pan, Nassim Oufattole, Wei-Hung Weng, Hanyi Fang, and Peter Szolovits. 2020. What Disease does this Patient Have? A Large-scale Open Domain Question Answering Dataset from Medical Exams. arXiv:2009.13081 https://arxiv.org/abs/2009.13081

    arXiv 2020

Show all 38 references
  1. [9]

    Joshi, Hanna Moazam, Heather Miller, Matei Zaharia, and Christopher Potts

    Omar Khattab, Arnav Singhvi, Paridhi Maheshwari, Zhiyuan Zhang, Keshav San- thanam, Sri Vardhamanan, Saiful Haq, Ashutosh Sharma, Thomas T. Joshi, Hanna Moazam, Heather Miller, Matei Zaharia, and Christopher Potts. 2023. DSPy: Compiling Declarative Language Model Calls into Se...

    2023 arXiv

  2. [10]

    Anthony Jethro Lieander, Hui Wang, and Karen Rafferty. 2025. Prompt Opti- mization with Two Gradients for Classification in Large Language Models.AI6, 8 (2025). doi:10.3390/ai6080182

    2025 doi

  3. [11]

    Rithesh Murthy, Ming Zhu, Liangwei Yang, Jielin Qiu, Juntao Tan, Shelby Hei- necke, Caiming Xiong, Silvio Savarese, and Huan Wang. 2025. Promptomatix: An Automatic Prompt Optimization Framework for Large Language Models. arXiv:2507.14241 https://arxiv.org/abs/2507.14241

    2025 arXiv

  4. [12]

    Ryan, Josh Purtell, David Broman, Christopher Potts, Matei Zaharia, and Omar Khattab

    Krista Opsahl-Ong, Michael J. Ryan, Josh Purtell, David Broman, Christopher Potts, Matei Zaharia, and Omar Khattab. 2024. Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programs. arXiv:2406.11695 https://arxiv.org/abs/2406.11695

    2024 arXiv

  5. [13]

    Joon Park, Kyohei Atarashi, Koh Takeuchi, and Hisashi Kashima. 2025. Emulating Retrieval Augmented Generation via Prompt Engineering for Enhanced Long Context Comprehension in LLMs. arXiv:2502.12462 https://arxiv.org/abs/2502. 12462

    2025 arXiv

  6. [14]

    Reid Pryzant, Dan Iter, Jerry Li, Yin Tat Lee, Chenguang Zhu, and Michael Zeng

  7. [15]

    João Rodrigues and António Branco. 2024. Meta-prompting Optimized Retrieval- augmented Generation. InProgress in Artificial Intelligence (EPIA 2024). doi:10. 1007/978-3-031-73503-5_17

    2024

  8. [16]

    Harsh Trivedi, Niranjan Balasubramanian, Tushar Khot, and Ashish Sabharwal

  9. [17]

    Guangzhi Xiong, Qiao Jin, Xiao Wang, Yin Fang, Haolin Liu, Yifan Yang, Fangyuan Chen, Zhixing Song, Dengyu Wang, Minjia Zhang, Zhiyong Lu, and Aidong Zhang. 2025. RAG-Gym: Systematic Optimization of Language Agents for Retrieval-Augmented Generation. arXiv:2502.13957 https://a...

    2025 arXiv

  10. [18]

    InProceedings of the 61st Annual Meeting of the Association for Computational Linguistics (ACL)

    Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge- Intensive Multi-Step Questions. InProceedings of the 61st Annual Meeting of the Association for Computational Linguistics (ACL). doi:10.18653/v1/2023.acl- long.557

    2023 doi

  11. [19]

    Cohen, Ruslan Salakhutdinov, and Christopher D

    Zhilin Yang, Peng Qi, Saizheng Zhang, Yoshua Bengio, William W. Cohen, Ruslan Salakhutdinov, and Christopher D. Manning. 2018. HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering. InProceedings of the 2018 Conference on Empirical Methods in Natural Langua...

    2018

  12. [20]

    Le, Denny Zhou, and Xinyun Chen

    Chengrun Yang, Xuezhi Wang, Yifeng Lu, Hanxiao Liu, Quoc V. Le, Denny Zhou, and Xinyun Chen. 2023. Large Language Models as Optimizers. arXiv:2309.03409 https://arxiv.org/abs/2309.03409

    2023 arXiv

  13. [22]

    Yongchao Zhou, Andrei Ioan Muresanu, Ziwen Han, Keiran Paster, Silviu Pitis, Harris Chan, and Jimmy Ba. 2022. Large Language Models Are Human-Level Sundararaman et al. Prompt Engineers. arXiv:2211.01910 https://arxiv.org/abs/2211.01910Appendix A: FLARE Iterative Improvement Pr...

    2022 arXiv

  14. [23]

    If no clear emotion is expressed, output ONLY`[27]`

  15. [24]

    ↩→Only label emotions that are clearly present from wording or ↩→context

    Do NOT add extra emotions just because they are compatible. ↩→Only label emotions that are clearly present from wording or ↩→context

  16. [25]

    Thanks, this is so exciting!

    Multi-label is allowed only when multiple distinct emotions ↩→are clearly expressed (e.g., "Thanks, this is so exciting!"→ ↩→[13,15])

  17. [26]

    lol" or

    Sarcasm: focus on the underlying attitude (mocking, hostile, supportive, etc.), not surface words like "lol" or "thanks"

  18. [27]

    The lake is steaming! The water is way warmer than the air

    If the text directly thanks someone, almost always include ↩→[15] (gratitude) unless clearly sarcastic or hostile. WHEN TO USE [27] NEUTRAL Use [27] when: - The text is informational, descriptive, or procedural with no ↩→clear emotional stance: "The lake is steaming! The water...

  19. [28]

    Decide which emotion IDs (0-27) are **explicitly** expressed

  20. [29]

    Output **only** a JSON-style list of integer IDs in **strict ascending order**

  21. [30]

    EVERY POST YOU MAKE IS AMAZING THANKS MY MAN

    Output format MUST be exactly one of these patterns (no extra spaces/characters): -`[15]` -`[0, 15]` -`[3, 4, 20]` -`[27]` Format details (must follow exactly): - Left bracket`[` - One or more integers from 0 to 27 - Sorted ascending - Separated by`,`(comma + single space) - R...

  22. [31]

    Decide which emotion IDs (0-27) are **explicitly** expressed ↩→in that text

  23. [32]

    Return ONLY a JSON-style list of integer IDs in STRICT ↩→ASCENDING ORDER

  24. [33]

    - Left bracket`[`then one or more integers in`0-27`, sorted ascending, separated by`,`(comma + single space), then right bracket`]`

    Output format MUST be exactly one of these shapes (no extra spaces/characters): -`[15]` -`[0, 15]` -`[3, 4, 20]` -`[27]` Format rules (non-negotiable): - Output = one line, one list. - Left bracket`[`then one or more integers in`0-27`, sorted ascending, separated by`,`(comma +...

  25. [34]

    LOL", ":)

    love). A2. Evidence-only reasoning - Use only what is **explicitly expressed**: words, emojis, punctuation, style (e.g., "LOL", ":)"). - Do not infer emotions just from events: - Bad/sad event described flatly→can still be`[27]`unless sadness/grief/etc. is explicit. - Good eve...

  26. [35]

    Decide the **minimal sufficient** set of emotion IDs based on ↩→the above rules

  27. [36]

    Sort them in ascending order

  28. [37]

    Output exactly one list in the form`[id]`or`[id1, id2, ↩→id3]`, with comma+space as the separator

  29. [38]

    Do **not** include any other characters, words, or ↩→explanation

  30. [2023]

    Gradient Descent

    Automatic Prompt Optimization with “Gradient Descent” and Beam Search. arXiv:2305.03495 https://arxiv.org/abs/2305.03495

    arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.