Disentangling Reasoning Logic to Resolve Explicit Knowledge Conflicts

Jiamou Liu; Kaiqi Zhao; Meng-Fen Chiang; Michael Witbrock; Xianda Zheng; Yuan Fang; Zijian Huang

REVIEW 5 major objections 6 minor 16 references

Disentangling Reasoning Logic to Resolve Explicit Knowledge Conflicts

T0 review · 5 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read The paper claims that by disentangling conflicting contexts into text and graph reasoning traces and training with verifiable logical-coherence and consistency rewards, KCR lets small open models adjudicate explicit knowledge conflicts…

desk verdict Worth a referee's time for the KCR framework and internal gains, but the headline '7B beats GPT-5.1' claim is unsupported because the comparison is uncontrolled. read the letter →

arxiv 2508.01273 v3 pith:ATWJQJGB submitted 2025-08-02 cs.AI

Xianda Zheng , Zijian Huang , Meng-Fen Chiang , Jiamou Liu , Yuan Fang , Michael Witbrock , Kaiqi Zhao This is my paper

classification cs.AI

keywords explicitknowledgeconflictsconflictreasoningtraceslocalgraphsreinforcementlearningwithverifiablerewardsGRPOlogiccoherencerewardretrieval-augmentedgeneration

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

The paper proposes KCR (Knowledge Conflict Reasoning), a framework that tries to establish that explicit knowledge conflicts—questions paired with two candidate answers and two contradictory supporting contexts—can be resolved reliably by structurally disentangling the underlying reasoning logic rather than by scaling the model or hand-crafting fixed fusion rules. It claims that the root difficulty is logic entanglement and case heterogeneity, and that converting conflicting contexts into discrete textual and graph reasoning traces, then training the backbone with reinforcement learning over verifiable rewards for logical coherence and consistency, instills a generalizable adjudication policy. On the popQA and strategyQA benchmarks the authors report that KCR lifts a Llama 3.2-3B backbone to 87.5% semantic accuracy on popQA and a Qwen 2.5-7B backbone to 86.67%, outperforming Astute RAG running on GPT-5.1, which reaches 66.94%. If this holds, conflict adjudication becomes trainable into small open models instead of depending on proprietary reasoning systems, and the method offers a template for reasoning-path supervision where no single gold step-level chain exists.

What carries the argument

The load-bearing object is the paired reasoning-trace representation: for each candidate answer, a set of textual traces $RT^{(T)}_i$ and a set of graph traces $RT^{(G)}_i$ are extracted, and these reference sets serve as the target structure the generator must align with. The training machinery is GRPO with a composite reward $R_{\text{coh}} + R_{\text{con}} + R_{\text{correct}}$, where the coherence term is a binary comparison of absolute differences in Jensen-Shannon divergence between the generated trace and the two reference sets, and the consistency term is a binary agreement check between the generated answer and generated trace on the same candidate side. The design choice of discrete rewards matters: the authors argue it sharpens group-relative advantage estimation during RLVR sampling and stabilizes convergence.

What would settle it

Re-run the three decoding baselines (CUAD, AdaCAD, CoCoA) independently on popQA and check whether their ACC_L, ACC_EM, and ACC_CEM values really are identical; if they remain numerically identical across distinct decoding strategies, the baseline column is not trustworthy. Then fine-tune a strong proprietary model (e.g., GPT-5.1) on the same training split and evaluate it on the same test split: if its ACC_L reaches or exceeds KCR's 87.5% on popQA, the claim of surpassing proprietary models reduces to a training-asymmetry effect.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is that a model's failure on explicit knowledge conflicts is a reasoning-structure failure, not a scale failure, and that this failure can be corrected by a two-phase procedure. First, KCR decomposes each candidate answer and its context into two complementary sets of reasoning traces: textual traces written as alternating entity-to-relation-to-entity chains, and graph traces taken as directed paths in a locally constructed knowledge graph that intersect the query entity or relation. Second, it trains the backbone as a policy with GRPO, scoring each rollout with a binary logic coherence reward that compares the Jensen-Shannon divergence of the generated reasoning chain against the correct and incorrect trace sets, plus a binary logic consistency reward that checks the generated answer and the generated trace side with the same candidate, using normalized Levenshtein similarity. The authors report that the trained 3B and 7B models substantially exceed both their backbones and the strongest proprietary baseline, with the discrete reward variants giving stable convergence and the coherence reward carrying the main improvement in the ablation.

Load-bearing premise

The load-bearing premise is that the performance comparison is fair: KCR is fine-tuned on the training split of popQA and strategyQA, whereas the GPT-4o and GPT-5.1 baselines it beats are only prompted, so the observed gap could reflect benchmark exposure rather than the disentanglement mechanism itself.

Editorial extensions

If this is right

If the central claim is right, explicit knowledge conflict adjudication no longer requires frontier closed-source models: 3B and 7B open backbones trained with KCR can surpass a GPT-5.1-based Astute RAG pipeline on the evaluated benchmarks.
The ablations imply that the logic coherence reward is the main source of improvement, which means the value of the method comes primarily from structurally comparing generated reasoning with correct versus incorrect trace sets, not from the consistency check alone.
The success of both text and graph trace variants across backbones suggests that the learned adjudication policy is not tied to a single surface format, and that graph traces can help even when the backbone was not explicitly trained on graph inputs.
Because the reward is verifiable at the process level rather than token level, the method applies to conflict scenarios where there is no single gold reasoning chain, a situation that rules out ordinary supervised step-level training.
KCR trained on popQA and strategyQA is reported to improve performance specifically on questions requiring stricter logical chains, so the method's benefit should concentrate on multi-step adjudication rather than trivial factual recall.

Reading between the lines

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

Editorial inference: the reported margin over GPT-5.1 may be inflated by an asymmetry in training exposure—KCR is fine-tuned on the benchmark's training split while the proprietary baselines are evaluated zero-shot or with prompting only—so a matched fine-tuned proprietary baseline could shrink the gap.
Editorial inference: the method's dependence on GPT-4o-mini for knowledge-graph construction and as judge introduces an API dependency; a fully open variant that builds graphs with an open extractor would test whether the graph traces themselves, rather than the proprietary extractor, cause the gain.
Editorial inference: the binary coherence reward compares the generated chain against pre-extracted trace sets, so the whole pipeline inherits the quality of those extractions; if the frozen Qwen2.5-7B or GPT-4o-mini extractor makes systematic errors, the reward can reinforce the wrong side.
Editorial inference: the dyadic two-candidate formulation leaves open whether the policy transfers to multi-party conflicts or partially correct sources; a direct extension would construct trace sets for three or more candidates and score coherence against each, testing whether the learned policy scales beyond binary choice.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

The paper is worth reading for its core idea: instead of asking a model to adjudicate conflicting contexts on the fly, KCR first disentangles each side into structured reasoning traces (textual chains and local knowledge graphs), then uses RLVR with coherence and consistency rewards to teach a reasoning policy. That combination is genuinely new, and the internal ablations in Figure 4 show the method does much better than SFT or vanilla GRPO on the same backbone. The authors also ship code and are candid about the computational cost and the dyadic/English-only scope. Credit where it's due: the framework is coherent and the internal gains look real.

The problems are in the headline comparisons. The abstract says a 7B KCR model outperforms GPT-4o and GPT-5.1, but those proprietary models are evaluated zero-shot while KCR is trained for 10 epochs on the training split of popQA and strategyQA. That conflates method quality with benchmark exposure. The stress-test note is right, and reading the paper doesn't dissolve it — Table 2 confirms the setup. Worse, the rows for CUAD, AdaCAD, and CoCoA (Default) are numerically identical on all metrics for both datasets. Three distinct decoding strategies producing exactly the same numbers is implausible; that looks like placeholder values, and it makes the "Overall Improvement vs SOTA" row unreliable.

There are also methodological soft spots in the rewards. The coherence reward in Eq. (4)-(6) compares the generated trace to the reference sets via the absolute difference of a smoothness measure (sum of JS divergences between adjacent steps), not a structural or semantic match. So a generated trace that is simply "smooth" can be rewarded even if its content is wrong. The consistency reward in Eq. (8) uses a "normalized Levenshtein similarity" that is never formally defined, and it compares a generated chain against a set of reference traces without saying how the set-level similarity is computed. And Eq. (6) explicitly uses ground-truth labels to decide which side is "correct" for the coherence reward, so the training signal partly re-encodes the answer labels into a reasoning-style reward. These are fixable, but they weaken the stated mechanism.

All that said, the central idea is plausible and the internal results are meaningful. I would send this to review, with a clear request: control the comparison (add a fine-tuned proprietary baseline, or evaluate KCR zero-shot), re-run the suspicious baseline rows, and define the similarity functions properly. A reader interested in knowledge conflicts or RLVR for reasoning will get value from the framework even if the headline claim needs to be walked back.

Referee Report

5 major / 6 minor

Summary. The paper proposes KCR (Knowledge Conflict Reasoning), a two-stage framework for explicit knowledge conflicts. In the first stage, it decomposes each candidate answer's supporting context into textual reasoning traces and local knowledge graphs. In the second stage, it trains a backbone LLM with GRPO using two reward signals: a logic coherence reward based on JS divergence between consecutive reasoning steps, and a logic consistency reward based on similarity between the generated trace/answer and candidate trace/answer sets. The authors evaluate on popQA and strategyQA against CAD, CUAD, AdaCAD, CoCoA, and Astute RAG, and report large gains over backbone models, claiming that KCR-enhanced 7B and 8B models outperform GPT-4o and GPT-5.1. The paper also reports ablations showing that both rewards contribute and provides qualitative cases of structured adjudication.

Significance. If the empirical results are taken at face value, KCR would be a meaningful advance for conflict adjudication: the hybrid text/graph trace disentanglement and the RLVR-style two-reward design are a principled attempt to separate valid from spurious reasoning paths. The paper releases code, provides detailed prompts, and is candid about limitations such as external LLM dependency, dyadic-only conflict scope, and English-only evaluation. However, the headline comparative claims are currently not supportable because the evaluation protocol does not match the training conditions of the proprietary baselines and because the baseline table contains an implausible artifact. The internal ablations and backbone-level improvements are plausible and interesting, but the central superiority claim needs a corrected and fairer comparison before it can be accepted.

major comments (5)

[Section 4.1, Table 2] The headline comparison against proprietary models is not apples-to-apples. KCR is trained for 10 epochs on the 8:1:1 training split of popQA and strategyQA, while the proprietary baselines (GPT-4o, GPT-5.1, and Astute RAG built on them) are evaluated zero-shot or with prompting only. The reported gaps, such as KCR with Llama 3.2-3B reaching 87.5% ACC_L on popQA versus 66.94% for Astute RAG (GPT-5.1), therefore conflate the method's contribution with supervised exposure to the benchmark's answer distribution and prompt format. The 'Overall Improvement vs SOTA' row should be recomputed after adding a fine-tuned or in-distribution proprietary baseline, or after evaluating a KCR variant that is not trained on the benchmark, and the paper should state the training/evaluation asymmetry explicitly.
[Table 2] The rows for CUAD (Default), ADACAD (Default), and COCOA (Default) report numerically identical values on all three metrics for both datasets (popQA: 0.3431/0.2958/0.3056; strategyQA: 0.5040 for all). Three distinct decoding strategies cannot plausibly produce identical results at this precision, which suggests placeholder or mislabeled values. These rows must be rerun and corrected, and the final 'Overall Improvement vs SOTA' row must be recomputed from the corrected values.
[Section 3.2.1, Eq. (6)] The coherence reward uses the ground-truth answer to choose which reference trace set is 'correct' ('Without loss of generality, let A1 denote the correct answer'). This is a legitimate training signal, but it means the reward is partly a re-encoding of answer labels into a reasoning-style signal. The paper should clearly separate this from the claim in Section 3.2.2 that 'ground-truth labels are not utilized' for the consistency reward. Moreover, Section 3.2.3 introduces an R_correct term in the cumulative reward without defining it; if that term is also label-based, the no-label claim is further weakened and the composition of the final reward needs to be specified precisely.
[Section 3.2.1, Eqs. (4)-(7)] The coherence score l(·) is a scalar sum of JS divergences between consecutive steps in a single trace, and Δl compares that scalar to the scalar of a reference trace set. This does not measure whether the generated trace's steps are structurally or semantically aligned with the correct trace; two logically unrelated chains can have identical aggregate divergence. The reward therefore does not implement the stated objective of 'approximat[ing] the logical structure present in valid contexts.' The authors should either define a trace-level alignment score or provide evidence that aggregate JS divergence correlates with trace-level correctness.
[Section 3.2.2, Eqs. (8)-(9)] Normalized Levenshtein similarity is defined for strings, but the reward compares 'sets' of reasoning traces RT_i as well as answer strings A_i. The manuscript does not define how set-level Levenshtein similarity is computed (e.g., maximum over pairs, average over pairs, or a sequence alignment over the set). The reward's behavior depends on this choice, so the definition must be specified and the chosen default justified with an ablation or example.

minor comments (6)

[Section 3.2.1, Eq. (4)] The notation is inconsistent: the summation limit uses len(R_i) while the text uses |R_i| for the number of steps; use one notation throughout.
[Section 3.2.3, Eqs. (10) and (12)] The symbol ε is used both for the numerical stability constant in the advantage normalization (Eq. 10) and for the GRPO clipping parameter (Eq. 12); use distinct symbols to avoid ambiguity.
[Section 3.2.3] The cumulative reward is defined as ri = R_coh + R_con + R_correct, but R_correct is never formally defined; define it explicitly and state whether it uses the ground-truth answer.
[Appendix A.2] There is a typo in the heading: 'Reprodicibility Statement' should be 'Reproducibility Statement'.
[Appendix B.1] The LLaMA answer extraction patterns list several alternative answer_patterns without explaining the order or conditions under which each is applied; clarify how the final answer string is selected.
[Figures 3 and 4] The figures would benefit from explicit axis labels and error bars, especially since the paper states that experimental results are averaged over three runs; currently the reported variability is not visible.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: KCR trains on labeled splits and evaluates on held-out test data; the baseline asymmetries and identical table rows are experimental-validity concerns, not derivation-circularity.

full rationale

The derivation chain is self-contained as a supervised/RL training procedure rather than a circular prediction. The rewards in Eqs. (6)-(9) explicitly depend on the ground-truth answer assignment ("let A1 denote the correct (ground-truth) answer") and on similarity to reference traces extracted from the contexts; this is ordinary label-based reward design, not a hidden re-encoding of the test outcome, because the model is trained on the 8:1:1 training split and evaluated on the held-out test split (Sec. 4.1). The reference trace sets and graph paths (Eqs. 1-3) are constructed from inputs (query, candidate answers, contexts) and do not presuppose the conclusion that KCR improves adjudication. No load-bearing self-citation appears: citations to ConflictQA, GRPO, GraphRAG, and decoding methods are external, and no uniqueness theorem from the authors' own prior work is invoked. The Limitations section honestly notes that verifiable rewards are less reliable without clear ground truth, which is a scope limitation rather than circularity. The Table 2 comparison is asymmetrical (KCR is fine-tuned on the benchmark training split while GPT-4o/GPT-5.1 baselines are not) and the identical CUAD/ADACAD/COCOA rows are implausible, but these are threats to the validity of the comparative claim, not instances of a claimed result reducing to its own inputs. Likewise, using GPT-4o-mini as both graph extractor and LLM-as-judge is a potential shared-bias measurement issue, not a definitional circularity. Accordingly, no specific circular step can be quoted and reduced by construction, so the circularity score is 0.

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

The central claim does not introduce new physical or mathematical entities, but it depends on several operational assumptions about what the extracted reasoning traces and similarity scores mean. The most fragile assumptions concern the validity of the LLM-as-judge evaluation and the interpretation of JS divergence and Levenshtein similarity as measures of reasoning logic.

free parameters (4)

number of training epochs = 10
Set to 10 across all experiments; no sensitivity analysis or justification is provided.
batch size = 32
Set to 32; no ablation is reported for this hyperparameter.
GRPO clipping coefficient epsilon
Appears in Eq. (12) but no numerical value is reported; a free hyperparameter in the objective.
GRPO group size G
The number of sampled outputs per group in GRPO is not stated; it affects reward normalization and policy gradient variance.

assumptions (5)

domain assumption LLM-as-judge (GPT-4o-mini) yields accurate semantic evaluation
Evaluation metric ACC_L depends on this judge; no systematic validation against human labels is reported (Section 4.1, Appendix B.5).
domain assumption Reasoning traces extracted by the frozen LLM and GPT-4o-mini faithfully represent the logic of the contexts
The reward design assumes the extracted traces are the correct disentangled logic; no human verification of trace quality is provided (Sections 3.1.2 and 3.1.3).
ad hoc to paper JS divergence of normalized embedding vectors measures logical coherence between reasoning steps
Equations (4)-(5) use this specific operationalization; no theoretical or empirical justification is given that lower JS divergence corresponds to better reasoning logic (Section 3.2.1).
ad hoc to paper Levenshtein similarity between string sets is a meaningful consistency measure
Equation (8) uses S(RT_i, RT_j) but RT_i are sets of traces, and the paper does not define how set-level string similarity is computed (Section 3.2.2).
domain assumption PopQA and strategyQA splits are representative of explicit knowledge conflicts
Only two English benchmarks are used; the paper's own Limitations section notes the restricted linguistic and benchmark scope.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Disentangling Reasoning Logic to Resolve Explicit Knowledge Conflicts." pith.science (2026). https://pith.science/paper/ATWJQJGB

@misc{pith2026250801273,
  author       = {Pith},
  title        = {Pith review of: Disentangling Reasoning Logic to Resolve Explicit Knowledge Conflicts},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ATWJQJGB}},
  note         = {Machine review of arXiv:2508.01273}
}

read the original abstract

Explicit knowledge conflicts, occurring when retrieved contexts contain contradictory information, pose a fundamental challenge for Large Language Models (LLMs) as they integrate increasingly diverse data sources. The core difficulty lies in the complexity of entangled narratives and heterogeneous conflict patterns, which frequently exceeds the reasoning capacity of standard backbone architectures. We propose \textbf{\textsc{Kcr}} (Knowledge Conflict Reasoning), a framework that adjudicates contradictions by systematically structuring their underlying logic. \textsc{Kcr} disentangles conflicting contexts into discrete sets of reasoning traces, utilizing a hybrid representation of text and graphs to facilitate systematic comprehension. It then employs a Reinforcement Learning with Verifiable Rewards (RLVR) paradigm to instill a reasoning policy that maximizes logical consistency while suppressing spurious paths derived from contradictory evidence. Extensive evaluations demonstrate that \textsc{Kcr} yields substantial performance gains. Notably, a 7B model enhanced by \textsc{Kcr} achieves adjudication capabilities that significantly outperform leading proprietary models, including GPT-4o and GPT-5.1, on complex tasks. Code is available at https://github.com/zhengxianda/KCR.

Figures

Figures reproduced from arXiv: 2508.01273 by the authors.

**Figure 2.** Overview of the KCR framework. Our approach structures contradictory evidence into intermediate reasoning traces, followed by a reinforcement learning stage designed to maximize logical consistency and eliminate spurious paths during conflict resolution. tion Extraction in other cases. Extraction prompt templates are detailed in Appendix B.2. 3.1.2 Textual Reasoning Traces With eq and rq extracted, the next step is … view at source ↗

**Figure 3.** Impact of Reasoning Coherence. Analysis of model performance across varying degrees of logical coherence (JS Divergence). Lower scores indicate stricter logical adherence [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Figures from the paper (1 more)

**Figure 4.** Figure 4: Ablation Study. KCR excels in scenarios requiring rigorous logical chains (popQA) compared to looser reasoning tasks. (b) validates that both Logic Coherence and Consistency rewards are essential for robust conflict adjudication. knowledge and context. 5.2 Explicit Kno…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

16 extracted references · 15 canonical work pages

[1]

Rimsky-Korsakov -> composer -> wrote -> Trombone Concerto \ newline

work page
[2]

Rimsky-Korsakov -> composer -> composed -> 1877 -> Trombone Concerto \newline

work page
[3]

Rimsky-Korsakov -> composer -> showcased -> 'virtuosic solo passages -> Trombone Concerto \ newline

work page
[4]

Rimsky-Korsakov -> composer -> showcased -> intricate orchestration -> Trombone Concerto \newline

work page
[5]

Rimsky-Korsakov -> composer -> wrote -> for -> Trombone and Military Band -> Trombone Concerto \newline

work page
[6]

Rimsky-Korsakov -> composer -> wrote -> in 1877 -> Trombone Concerto \newline

work page
[7]

Leonov -> fellow marine officer -> commissioned -> Trombone Concerto \newline

work page
[8]

Concerto for Trombone and Military Band,

Rimsky-Korsakov -> composer -> premiered -> at Kronstadt -> Trombone Concerto \newline \newline Conflicting path set 2: \newline Johann Georg Albrechtsberger -> composer -> Trombone Concerto \ newline Johann Georg Albrechtsberger -> composer -> variety of works -> Trombone Concerto \newline Johann Georg Albrechtsberger -> composer -> late 1700s -> Trombon...

work page

Show all 16 references

[11]

The first answer attributes the Trombone Concerto to Johann Georg Albrechtsberger.\newline
[12]

The second answer attributes the Trombone Concerto to Nikolai Rimsky-Korsakov.\newline
[13]

Both answers provide context about the composer and the piece, suggesting they have different composers.\newline
[14]

This time period does not overlap.\newline

Albrechtsberger lived from 1736 to 1809, while Rimsky-Korsakov lived from 1844 to 1908. This time period does not overlap.\newline

1908
[15]

The context provided for the second answer mentions that the Trombone Concerto was composed in 1877, which aligns with Rimsky-Korsakov' s lifetime.\newline
[16]

The first answer provides no specific year or context about when the Trombone Concerto was composed.\newline </think>\newline <answer>\newline Rimsky-Korsakov\newline </answer>
[2023]

<think>(.*?)</ think>

Merging generated and retrieved knowledge for open-domain QA. InEMNLP, pages 4710–4728. Zheng Zhao, Emilio Monti, Jens Lehmann, and Haytham Assem. 2024. Enhancing contextual un- derstanding in large language models through con- trastive decoding. InNAACL, pages 4225–4237. Lian...

2024
[2025]

Solomon Kullback and Richard A Leibler

Cocoa: Confidence and context-aware adap- tive decoding for resolving knowledge conflicts in large language models.CoRR, abs/2508.17670. Solomon Kullback and Richard A Leibler. 1951. On information and sufficiency.The annals of mathe- matical statistics, 22(1):79–86. Kyungjae ...

1951 arXiv

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [1]

Rimsky-Korsakov -> composer -> wrote -> Trombone Concerto \ newline

work page

[2] [2]

Rimsky-Korsakov -> composer -> composed -> 1877 -> Trombone Concerto \newline

work page

[3] [3]

Rimsky-Korsakov -> composer -> showcased -> 'virtuosic solo passages -> Trombone Concerto \ newline

work page

[4] [4]

Rimsky-Korsakov -> composer -> showcased -> intricate orchestration -> Trombone Concerto \newline

work page

[5] [5]

Rimsky-Korsakov -> composer -> wrote -> for -> Trombone and Military Band -> Trombone Concerto \newline

work page

[6] [6]

Rimsky-Korsakov -> composer -> wrote -> in 1877 -> Trombone Concerto \newline

work page

[7] [7]

Leonov -> fellow marine officer -> commissioned -> Trombone Concerto \newline

work page

[8] [8]

Concerto for Trombone and Military Band,

Rimsky-Korsakov -> composer -> premiered -> at Kronstadt -> Trombone Concerto \newline \newline Conflicting path set 2: \newline Johann Georg Albrechtsberger -> composer -> Trombone Concerto \ newline Johann Georg Albrechtsberger -> composer -> variety of works -> Trombone Concerto \newline Johann Georg Albrechtsberger -> composer -> late 1700s -> Trombon...

work page