Pith. sign in

REVIEW 4 major objections 6 minor 14 references

Coupling retrieval, memory, and teaching control as one evolving process raises a DSP tutor's judged score from 5.32 to 9.23.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

A hierarchical multi-agent tutoring system that jointly evolves a student profile, a value-scored knowledge base, and a teaching policy reports large LLM-judged gains over static RAG and single-agent baselines.

T0 review reviewed 2026-08-03 challenge →

load-bearing objection CogEvo-Edu is a coherent three-layer tutoring architecture with sensible ablations, but the headline 5.32-to-9.23 margin is not credible because it rests entirely on an unvalidated LLM judge ensemble scoring simulated students. the 4 major comments →

arxiv 2512.00331 v1 pith:YPVTWDTV submitted 2025-11-29 cs.AI cs.MA

CogEvo-Edu: Cognitive Evolution Educational Multi-Agent Collaborative System

Yefeng Wu , Yuchen Song , Yecheng Zhao , Ling Wu , Shan Wan This is my paper
classification cs.AI cs.MA
keywords cognitive evolutionmulti-agent systemretrieval-augmented generationstudent modelingknowledge evolutionLLM-as-a-judgedigital signal processing tutoringhierarchical reinforcement learning
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The reading

Most LLM tutors pair a single model with a static retrieval pipeline, which fails over long dialogues. The paper argues that retrieval, memory, and teaching control must be treated as coupled, evolving processes, and instantiates this in CogEvo-Edu: a perception layer builds confidence-weighted student profiles, a knowledge layer scores chunks by frequency, recency, and semantic density to activate, compress, or forget them, and a meta-control layer orchestrates specialist agents while adapting the lower layers' hyperparameters. On DSP-EduBench, a new digital signal processing benchmark with heterogeneous resources and simulated students, the full system raises the average judge score from 5.32 to 9.23 and improves all six indicators over static RAG, simple memory, and a single-agent variant. The intended lesson is that joint co-evolution—not any single component—is what drives large, consistent gains in long-horizon tutoring.

Core claim

The paper's claim is that retrieval, memory, and control should be modeled as one coupled cognitive evolution process, and that doing so yields large gains in long-horizon tutoring. In CogEvo-Edu, a perception layer consolidates dual memories into confidence-weighted student profiles, a knowledge layer scores each chunk by interaction frequency, temporal decay, and semantic density to decide whether to keep, compress, or delete it, and a meta-control layer selects among specialist agents and adapts the lower layers' hyperparameters via inner and outer loops. On DSP-EduBench—a new DSP tutoring benchmark with heterogeneous resources and simulated students—the full system raises the average LLM

What carries the argument

The load-bearing mechanism is the spatiotemporal value function V(c_i) = α·(frequency norm) + β·exp(-Δt/τ) + γ·(semantic density), which partitions the knowledge base into active, soluble (semantically compressed), and deleted sets, making retrieval adaptive to use and recency. It is coupled with the CPL's confidence-weighted consolidation operator, which updates profile features with a momentum rule that reinforces or corrects based on semantic agreement, and the MCL's dual inner–outer loop, which formulates tutoring as hierarchical sequential decision making and supposedly adapts the CPL/KEL hyperparameters (η, α, β, γ, thresholds, τ) against a long-term learning objective.

Load-bearing premise

The entire empirical case depends on a three-model LLM-as-a-Judge ensemble scoring a self-built benchmark with simulated students and no human raters; if these judges reward fluent, profile-referential responses rather than real learning, the 5.32-to-9.23 margin may not reflect actual tutoring quality.

What would settle it

Have a panel of human DSP instructors rate the same interaction transcripts without knowing which system produced them; if human ratings do not place CogEvo-Edu clearly above Static RAG and the Single Agent variant, the headline margin is an artifact of the judge models. Alternatively, run the full system with the MCL outer loop disabled (fixed hyperparameters) and compare scores: no change would falsify the claim that hyperparameter adaptation contributes to the result.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Tutoring systems should replace fixed top-k retrieval with value-driven knowledge lifecycles that compress or delete content based on interaction frequency, recency, and semantic centrality.
  • Structured, confidence-weighted student profiles outperform sliding-window summaries for maintaining personalization over long dialogues.
  • A meta-control layer that dynamically selects specialist agents and adjusts difficulty and retrieval scope yields large gains in guidance quality and strategy flexibility.
  • DSP-EduBench provides a reusable vertical benchmark for evaluating tutors in signal processing and similar technical domains.
  • The largest gains come from coupling all three processes; each individual component (RAG, memory, single-agent with CPL+KEL) lags behind the full system.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same coupled-evolution design should transfer to other long-horizon technical tutoring domains (e.g., control theory, circuit analysis, statistical inference), but the magnitude of the gain may depend on how heterogeneous the knowledge resources are.
  • A natural next test is to replace the LLM-as-a-Judge ensemble with human expert ratings or a downstream learning measure (pre/post test on simulated students); if the margin shrinks, part of the 9.23 may reflect judge preference for fluent, profile-referential style rather than actual learning.
  • The paper describes the MCL outer loop as adapting CPL/KEL hyperparameters, but reports no experiment showing adaptation occurring; comparing the full dual-loop system against a version with fixed hand-set hyperparameters would isolate whether joint hyperparameter evolution is responsible for the gains.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper presents CogEvo-Edu, a hierarchical multi-agent tutoring system for digital signal processing (DSP). The architecture comprises a Cognitive Perception Layer (CPL) maintaining dual short-term and long-term memories with confidence-weighted consolidation; a Knowledge Evolution Layer (KEL) assigning spatiotemporal value scores to knowledge chunks to drive activation, compression, and forgetting; and a Meta-Control Layer (MCL) that orchestrates teaching agents and claims to adapt CPL/KEL hyperparameters via dual inner-outer loops. To evaluate, the authors construct DSP-EduBench, a vertical benchmark with heterogeneous knowledge sources, simulated student profiles, and long-horizon interaction scripts, and use a three-model LLM-as-a-Judge ensemble (GLM-4.5, DeepSeek-V3.1, Qwen3-max) to score six indicators. The full system is reported to improve over vanilla LLM, static RAG, simple memory, and a single-agent variant, raising the overall score from 5.32 to 9.23.

Significance. If the empirical claim were established, the paper would make a useful architectural contribution by coupling retrieval, memory, and teaching control. The CPL/KEL/MCL decomposition is clearly specified, and the layer-wise ablations are logically structured. DSP-EduBench, if released and validated, could help benchmark long-horizon tutoring systems. However, the headline margin currently rests on an unvalidated LLM-as-a-Judge instrument: there are no human ratings, no real-learner studies, no inter-judge agreement or error bars, and the judge pool includes a same-family model as the tutor backbone. There is also no experimental evidence that the MCL outer-loop adaptation actually occurred, despite this being part of the central 'joint evolution' claim. The contribution is promising but not yet established.

major comments (4)
  1. [§IV-B and §V-B (Table I)] The entire empirical case rests on a three-model LLM-as-a-Judge ensemble scoring simulated dialogues, with no human evaluation, no real-learner study, no inter-judge agreement, and no validation that judge scores correlate with learning outcomes. Because the tutor backbone is Qwen3-14B and one judge is Qwen3-max, a same-family bias is plausible. Moreover, §II-C and §III-C state that MCL's long-term objective is estimated by an LLM-as-a-Judge ensemble on DSP-EduBench, while Table I uses the same jury design; the optimization target and evaluation metric are therefore not independent. The abstract's central claim ('raises the overall score from 5.32 to 9.23...') needs a bias check, e.g., hold-out human ratings, judge/tutor family swap, or per-judge score reporting.
  2. [§III-C vs §V] The paper claims the MCL 'periodically aggregates interaction trajectories and estimates J(θ, λ)' and performs a 'joint update on θ and λ,' but Section V reports no experiment showing that θ or λ actually changed. Configurations (a)–(e) are described by fixed components, not by adapted hyperparameter values. If the reported runs used hand-set α, β, γ, τ_decay, θ_solid, θ_forget, η, τ_match, then the 'jointly evolving ... teaching policies' claim is unsupported. Please show hyperparameter trajectories, a before/after comparison, or clearly specify which parameters were updated and when.
  3. [§V-B, Table I] All scores are single-run point estimates without error bars, confidence intervals, or significance tests. The margin between Simple Memory (6.45) and Single Agent (6.42) is within rounding, and the conclusion that 'the largest gains arise when retrieval, memory, and control are treated as a coupled process' depends on both differences being reliable. Report multiple independent runs or bootstrap over scripts, with significance tests, and justify the score aggregation across turns.
  4. [§IV-A] DSP-EduBench is author-constructed with simulated student profiles and author-annotated 'ideal pedagogical strategy' ground truth. No annotation protocol, inter-annotator agreement, or public release is provided, and the simulated profiles/scripts are not validated against real DSP tutoring interactions. This makes absolute scores hard to interpret and the benchmark impossible to reproduce. Please release the benchmark and annotation guidelines, or at least report annotation consistency.
minor comments (6)
  1. [Introduction] Typo: 'viacognitive' is missing a space. Also use 'cognitive evolution' consistently rather than sometimes capitalizing it.
  2. [§III-A] The notation P^t = {(k_j, v_j, ω_j)} and Ψ(P^t, H^t) = P^t ⊕ LLM extract(H^t) is ambiguous: the fusion operator ⊕ is not precisely defined (how is semantic similarity computed, how is contradiction detected, how is τ_match used). Please give a concrete algorithmic description.
  3. [§III-B] The semantic density proxy D_sem(c_i) uses average cosine similarity with k-NN as a measure of pedagogical importance. This is an unexplained modeling assumption, and the parameters α, β, γ, τ_decay, θ_solid, θ_forget are introduced without any sensitivity analysis or ablation.
  4. [Figure 1] Figure 1 is not described in the text. Add a caption or a paragraph that explains the data flows among CPL, KEL, MCL, and the simulated student to help readers parse the architecture.
  5. [Table I] Dimension names are inconsistent across the text, radar chart, and table ('Knowledge Guidance' vs 'Knowledge-driven Capability', 'Strategy Flexibility' vs 'Strategy Switching Flexibility'). Also, no number of turns, scripts, or profiles per configuration is given.
  6. [References] The paper cites LLM-as-a-Judge work but does not discuss known judge biases (e.g., position bias, verbosity bias, self-preference) in the context of this evaluation. Consider adding a limitation paragraph.

Circularity Check

1 steps flagged

The system's optimization target and the reported evaluation metric are the same LLM-as-a-Judge ensemble on the same self-built benchmark.

specific steps
  1. fitted input called prediction [Section II-C (Related Work) and Abstract / Section IV-B]
    "CogEvo-Edu’s Meta-Control Layer combines these ideas: it orchestrates multiple teaching agents through an inner-loop policy and uses a long-term objective, estimated by an LLM-as-a-Judge ensemble on DSP-EduBench, to update both this policy and the hyperparameters governing CPL and KEL, turning orchestration and memory/knowledge evolution into a unified meta-control problem. ... Using a three-model LLM-as-a-Judge ensemble, CogEvo-Edu raises the overall score from 5.32 to 9.23 and improves all six indicators over static RAG, simple memory, and a single-agent variant."

    The paper states that the MCL's long-term objective is estimated by an LLM-as-a-Judge ensemble on DSP-EduBench, and then reports scores from the same type of ensemble on the same benchmark as evidence for the architecture. The optimization target and the evaluation metric are therefore the same construction: the 5.32-to-9.23 improvement is, at least in part, the system reporting the value of the objective it was designed to maximize. Without an external benchmark, a held-out judge, or human assessment, the evaluation loop is closed by definition and does not independently validate the cognitive-evolution claim.

full rationale

The central empirical claim—that jointly evolving student profiles, knowledge bases, and teaching policies yields large gains—is supported only by the authors' own DSP-EduBench and LLM-as-a-Judge ensemble. The circularity is not merely that the benchmark is self-constructed; it is that Section II-C explicitly describes the MCL's long-term objective as 'estimated by an LLM-as-a-Judge ensemble on DSP-EduBench,' while Section IV-B uses the same three-model jury to score the six indicators and the abstract reports the resulting 5.32-to-9.23 margin. Thus the reported superiority is partly a measure of how well the system minimized/maximized that specific jury's preferences, rather than an independent test. The absence of human ratings, external benchmarks, or held-out judges makes the loop unbroken. I also note that the MCL outer-loop adaptation is described but never experimentally demonstrated: no experiment shows hyperparameters actually changing, so the 'jointly evolving policies' claim is unsupported on an additional, non-circular ground. The Qwen3-max judge sharing a family with the Qwen3-14B tutor is a further validity risk but is a measurement-bias issue rather than a constructional circularity. No self-citation or uniqueness-theorem circularity is present. Overall, partial circularity: the objective used for system adaptation is the same as the evaluation metric, weighting the central result toward a score of 6.

Axiom & Free-Parameter Ledger

7 free parameters · 7 axioms · 0 invented entities

The reported gains are measured exclusively through a self-built benchmark and an unvalidated LLM judge, and every KEL/CPL control parameter is undisclosed, so the architecture's behavior cannot be audited or transferred without contacting the authors. No new physical or formal entities are introduced: 'spatiotemporal value' and 'cognitive evolution' are framings of scoring functions and control loops, not independent entities.

free parameters (7)
  • α, β, γ — KEL value-function weights = not reported
    Coefficients in V(c_i)=α·f_norm+β·exp(−Δt/τ)+γ·D_sem (Section III-B); no values or tuning procedure are given.
  • τ_decay — forgetting half-life = not reported
    Time-decay constant in the KEL value function (Section III-B); sets how quickly unused chunks lose value.
  • θ_solid, θ_forget — activation/deletion thresholds = not reported
    Partition the knowledge base into K_act/K_sol/K_del (Section III-B); directly control pruning aggressiveness.
  • η — profile consolidation learning rate = not reported
    Momentum step in the confidence update ω_new (Section III-A); determines how fast student-profile confidence changes.
  • τ_match — semantic match threshold = not reported
    Similarity threshold that classifies new evidence as reinforcement vs correction (Section III-A).
  • w — short-term memory window = not reported
    Size of the sliding window H_t that triggers consolidation (Section III-A).
  • k (semantic density neighbors); retrieval top-k = not reported
    Neighborhood size for D_sem and the fixed top-k used by the static-RAG baseline (Section III-B, V-A).
axioms (7)
  • domain assumption LLM-extracted features and the semantic-similarity match rule (τ_match) yield consistent, correct student-profile updates rather than amplifying errors.
    The consolidation operator Ψ (Section III-A) trusts LLM extraction and cosine-similarity matches to reinforce/correct long-term features; no error-propagation analysis is given.
  • domain assumption Exponential time decay with half-life τ_decay models the pedagogical value of knowledge chunks.
    V(c_i) includes exp(−Δt_i/τ_decay) (Section III-B) with no validation of the learning-science basis; no spacing-effect literature is cited.
  • ad hoc to paper Average cosine similarity to k-nearest neighbors is a valid proxy for a chunk's semantic density and pedagogical importance.
    D_sem is defined as mean cosine similarity over k-NN (Section III-B); no evidence that this tracks teaching value.
  • domain assumption LLM summarization (T_compress) preserves the information needed for tutoring after compression.
    Compressed chunks replace full indices for K_sol (Section III-B); information loss is asserted negligible without measurement.
  • domain assumption LLM-as-a-Judge ensemble scores are valid, unbiased measures of the six tutoring indicators.
    The judge pipeline (Section IV-B) is the sole outcome measure; no human-judge correlation or bias analysis is reported, and Qwen3-max shares a family with the Qwen3-14B tutor.
  • domain assumption Simulated student profiles (UserA/UserB/UserC) adequately represent real student cognition for comparing tutoring systems.
    DSP-EduBench (Section IV-A) is built entirely on simulated profiles; no validation against real learner data.
  • domain assumption The reward r_t and long-term objective J(θ,λ) can be estimated from mastery improvement, error-pattern extinction, and retention.
    MCL outer loop (Section III-C) treats J as the optimization target; how these quantities are measured is unspecified and never demonstrated in the experiments.

reviewed 2026-08-03 · how reviews work

0 comments
Cite this review

Pith. "Pith review of CogEvo-Edu: Cognitive Evolution Educational Multi-Agent Collaborative System." pith.science (2026). https://pith.science/paper/YPVTWDTV

@misc{pith2026251200331,
  author       = {Pith},
  title        = {Pith review of: CogEvo-Edu: Cognitive Evolution Educational Multi-Agent Collaborative System},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YPVTWDTV}},
  note         = {Machine review of arXiv:2512.00331}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Large language models (LLMs) are increasingly deployed as conversational tutors in STEM education, yet most systems still rely on a single LLM with a static retrieval-augmented generation (RAG) pipeline over course materials. This design struggles in complex domains such as digital signal processing (DSP), where tutors must maintain coherent long-term student models, manage heterogeneous knowledge bases, and adapt teaching strategies over extended interactions. We argue that retrieval, memory, and control should be treated as a coupled cognitive evolution process. We instantiate this view in CogEvo-Edu, a hierarchical educational multi-agent system comprising a Cognitive Perception Layer (CPL), a Knowledge Evolution Layer (KEL), and a Meta-Control Layer (MCL). CPL maintains dual memories and performs confidence-weighted consolidation to build structured, self-correcting student profiles under limited context. KEL assigns each knowledge chunk a spatiotemporal value that drives activation, semantic compression, and forgetting. MCL formulates tutoring as hierarchical sequential decision making, orchestrating specialized agents and jointly adapting CPL/KEL hyperparameters via a dual inner--outer loop. To evaluate CogEvo-Edu, we construct DSP-EduBench, a vertical benchmark for DSP tutoring with heterogeneous resources, simulated student profiles, and long-horizon interaction scripts. Using a three-model LLM-as-a-Judge ensemble, CogEvo-Edu raises the overall score from 5.32 to 9.23 and improves all six indicators over static RAG, simple memory, and a single-agent variant, demonstrating the value of jointly evolving student profiles, knowledge bases, and teaching policies.

Figures

Figures reproduced from arXiv: 2512.00331 by Ling Wu, Shan Wan, Yecheng Zhao, Yefeng Wu, Yuchen Song.

Figure 1
Figure 1. Figure 1: CogEvo-Edu: Cognitive Evolution Educational Multi-Agent Collaborative System [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Evaluation pipeline based on LLM-as-a-Judge principles. [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Comparative Evaluation Results Visualization Radar Chart [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

14 extracted references · 2 canonical work pages

  1. [1]

    Generative large language models for dialog-based tutoring: An early consideration of opportunities and concerns,

    B. D. Nye, D. Mee, and M. G. Core, “Generative large language models for dialog-based tutoring: An early consideration of opportunities and concerns,” inProceedings of the Workshop on Empowering Education with LLMs – the Next-Gen Interface and Content Generation, ser. CEUR Workshop Proceedings, vol. 3487. CEUR-WS.org, 2023, pp. 78–88. [Online]. Available:...

    2023

  2. [2]

    Advancing generative intelligent tutoring systems with GPT-4: Design, evaluation, and a modular framework for future learning platforms,

    S. Liu, X. Guo, X. Hu, and X. Zhao, “Advancing generative intelligent tutoring systems with GPT-4: Design, evaluation, and a modular framework for future learning platforms,” Electronics, vol. 13, no. 24, p. 4876, 2024. [Online]. Available: https://doi.org/10.3390/electronics13244876

    work page doi:10.3390/electronics13244876 2024

  3. [3]

    Retrieval-augmented generation for knowledge-intensive NLP tasks,

    P. Lewis, E. Perez, A. Piktus, F. Petroni, V . Karpukhin, N. Goyal, H. K ¨uttler, M. Lewis, W. Yih, T. Rockt ¨aschel, S. Riedel, and D. Kiela, “Retrieval-augmented generation for knowledge-intensive NLP tasks,” inAdvances in Neural Information Processing Systems, vol. 33, 2020, pp. 9459–9474. [Online]. Available: https://arxiv.org/abs/2005.11401

    Pith/arXiv arXiv 2020

  4. [4]

    Lpitutor: an LLM based personalized intelligent tutoring system using RAG and prompt engineering,

    Z. Liu, P. Agrawal, S. Singhal, V . Madaan, M. Kumar, and P. K. Verma, “Lpitutor: an LLM based personalized intelligent tutoring system using RAG and prompt engineering,”PeerJ Computer Science, vol. 11, p. e2991, 2025. [Online]. Available: https://doi.org/10.7717/peerj-cs.2991

    work page doi:10.7717/peerj-cs.2991 2025

  5. [5]

    Memorybank: Enhancing large language models with long-term memory,

    W. Zhong, L. Guo, Q. Gao, H. Ye, and Y . Wang, “Memorybank: Enhancing large language models with long-term memory,” in Proceedings of the AAAI Conference on Artificial Intelligence, vol. 38, no. 17. AAAI Press, 2024, pp. 19 724–19 731. [Online]. Available: https://doi.org/10.1609/aaai.v38i17.29946

    doi:10.1609/aaai.v38i17.29946 2024

  6. [6]

    Memgpt: Towards LLMs as operating systems,

    C. Packer, S. Wooders, K. Lin, V . Fang, S. G. Patil, I. Stoica, and J. E. Gonzalez, “Memgpt: Towards LLMs as operating systems,” 2024. [Online]. Available: https://arxiv.org/abs/2310.08560

    Pith/arXiv arXiv 2024

  7. [7]

    Evaluating very long-term conversational memory of LLM agents,

    A. Maharana, D.-H. Lee, S. Tulyakov, M. Bansal, F. Barbieri, and Y . Fang, “Evaluating very long-term conversational memory of LLM agents,” inProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Association for Computational Linguistics, 2024, pp. 13 851–13 870. [Online]. Available: https://doi.o...

    doi:10.18653/v1/2024.acl-long.747 2024

  8. [8]

    DRAGIN: Dynamic retrieval augmented generation based on the information needs of large language models,

    W. Su, Y . Tang, Q. Ai, Z. Wu, and Y . Liu, “DRAGIN: Dynamic retrieval augmented generation based on the information needs of large language models,” 2024. [Online]. Available: https://arxiv.org/abs/2403.10081

    Pith/arXiv arXiv 2024

  9. [9]

    Autogen: Enabling next-gen LLM applications via multi-agent conversation,

    Q. Wu, G. Bansal, J. Zhang, Y . Wu, B. Li, E. Zhu, L. Jiang, X. Zhang, S. Zhang, J. Liu, A. H. Awadallah, R. W. White, D. Burger, and C. Wang, “Autogen: Enabling next-gen LLM applications via multi-agent conversation,” 2023. [Online]. Available: https://arxiv.org/abs/2308.08155

    Pith/arXiv arXiv 2023

  10. [10]

    CAMEL: Communicative agents for “mind

    G. Li, H. A. A. K. Hammoud, H. Itani, D. Khizbullin, and B. Ghanem, “CAMEL: Communicative agents for “mind” exploration of large language model society,” 2023. [Online]. Available: https://arxiv.org/abs/2303.17760

    Pith/arXiv arXiv 2023

  11. [11]

    Knowledge tracing: Modeling the acquisition of procedural knowledge,

    A. T. Corbett and J. R. Anderson, “Knowledge tracing: Modeling the acquisition of procedural knowledge,”User Modeling and User-Adapted Interaction, vol. 4, no. 4, pp. 253–278, 1994. [Online]. Available: https://doi.org/10.1007/BF01099821

    doi:10.1007/bf01099821 1994

  12. [12]

    F. M. Lord and M. R. Novick,Statistical Theories of Mental Test Scores. Reading, MA: Addison-Wesley, 1968

    1968

  13. [13]

    Educationq: Evaluating LLMs’ teaching capabilities through multi-agent dialogue framework,

    Y . Shi, R. Liang, and Y . Xu, “Educationq: Evaluating LLMs’ teaching capabilities through multi-agent dialogue framework,” in Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Association for Computational Linguistics, 2025, pp. 32 799–32 828. [Online]. Available: https://doi.org/10.18653/v1/...

    doi:10.18653/v1/2025.acl-long.1576 2025

  14. [14]

    From generation to judgment: Opportunities and challenges of LLM-as-a-judge,

    D. Li, B. Jiang, L. Huang, A. Beigi, C. Zhao, Z. Tan, A. Bhattacharjee, Y . Jiang, C. Chen, T. Wu, K. Shu, L. Cheng, and H. Liu, “From generation to judgment: Opportunities and challenges of LLM-as-a-judge,” inProceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics, 2025, pp. 2757–2...

    doi:10.18653/v1/2025.emnlp-main.138 2025

This paper was first reviewed by deepseek-v4-flash on August 3, 2026.