Papers › ScholarSum: Student-Teacher Abstractive Summarization via Knowledge Graph Reasoning...

ScholarSum: Student-Teacher Abstractive Summarization via Knowledge Graph Reasoning and Reflective Refinement

17 Jun 2026arXiv:2606.18850added by Syntology

Bohou Zhang, Xiaoyu Tao, Mingyue Cheng, Huijie Liu, Qi Liu

Title, abstract, authors and date from arXiv's metadata (CC0); this paper is not in the Papers with Code archive (frozen 2025-07-28).

Abstractive summarization plays a crucial role in enabling efficient understanding of scientific literature, yet it inherently demands both linguistic fluency and factual faithfulness. Existing approaches often fail to reconcile these two requirements. Extractive methods rely on rigid sentence splicing that disrupts macro-level logical coherence, while large language model (LLM)-based generative approaches, despite mastering linguistic fluency, exhibit limited factual consistency. In this work, we propose ScholarSum, a hierarchical reflective graph-based framework that emulates a student-teacher writing process for fluent and faithful scientific summarization. ScholarSum first organizes the document into a hierarchical knowledge graph by segmenting it into semantically coherent units, whose multi-layered community structure captures global logic and macro-level themes. Guided by this global structure, the student generates an initial draft, which is subsequently refined through fine-grained evidence retrieval. To ensure factual consistency, a teacher-like reviewer then iteratively examines the draft, identifies unsupported content, and prompts targeted re-retrieval and rewriting until the summary meets rigorous quality standards. Extensive experiments demonstrate that ScholarSum significantly outperforms previous baselines in terms of both completeness and faithfulness. Our code is available at https://github.com/Xiaoyu-Tao/ScholarSum.

PaperPDFCode Syntology ranMethod as code

In Syntology View this paper on Syntology, its page in Syntology's graph. That page adds which of its functions are also present in other papers' code Syntology holds, with dates, and its dated run record as JSON and a README badge. It also lists the repositories linked to the paper, the abstract and the calls for agents.

Open this paper in Syntology's Atlas, the map of the papers in Syntology's graph and their citations.

For agents, on Syntology's MCP service (how to connect):

Code

Syntology 40 code samples harvested from 1 repository linked to this paper: 31 ran (of which 12 constructed an object rather than computing a result; 16 with no instrument failure: 1 honoured, 3 violated, 12 with no contract checked; 15 where Syntology's instrument failed) · 9 unverified. Where Syntology's instrument failed: 15 where our contract draft was wrong; failures of Syntology's instrument, not of the code.

By repository: found in paper text by Syntology: 40 samples from 1 repository, 31 ran (16 with no instrument failure; 15 where Syntology's instrument failed). The run record, sample by sample. “Ran” means executed on a synthesized input, not that the code is correct or reproduces the paper.

Levels written on 2026-09-22 (UTC): the dates Syntology's graph wrote each sample's verification level, not when the code executed.

Xiaoyu-Tao/ScholarSum found in paper text by SyntologySyntology: 40 samples, 31 ran (15 instrument-failed) · 9 unverified report

Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

40 samples harvested: 31 ran (of which 12 constructed an object rather than computing a result; 16 with no instrument failure: 1 honoured, 3 violated, 12 with no contract checked; 15 where Syntology's instrument failed) · 9 unverified. “Where Syntology's instrument failed” counts failures of Syntology's instrument, not of the code. Syntology's record for this page last changed ; when this build read Syntology's graph is in the build record. That date is not when the samples ran.

1ran · honoured contract
3ran · violated contract
15ran · our draft was wrong
12ran
9unverified

Licence: 0 of the 40 samples are pointer only, meaning Syntology does not serve that copy's text. This page shows no code text for any sample; each one links to its file in the repository.

Harvested from Xiaoyu-Tao/ScholarSum. “Ran” means the sample executed on a synthesized input. It does not mean the output is correct, and nothing here reproduces the paper's results. A line beginning “fixture:” under a sample that ran gives the arguments Syntology's fixture called the sample with, each as name:type/rank/element kind (? where none was recorded), and the type of what the fixture's run returned, exactly as the harness recorded them; the arguments and the returned type were recorded by two separate runs of the same fixture. No output value is shown, and a returned type says nothing about whether the output was right. “Level recorded” gives the UTC date when Syntology's graph wrote the sample's verification level, not when the code executed; where the level was written before Syntology's graph kept that date, the row says “level recorded, date not recorded”. “Honoured” and “violated” refer to a contract Syntology drafted from the code itself. “Our draft was wrong” and “the contract check could not run” are failures of Syntology's instrument, not of the code: the sample ran on Syntology's synthesized input, and then the drafted contract did not describe it, or the check itself raised an error (RAISES), lacked a dependency (DEP_MISSING) or ran out of time (TIMEOUT); hover the label for which. A metamorphic tier, where shown, is evidence of the output's shape, not that the sample implements the method.

Each sample ends with its code_sha256. A sample's code_sha256 is Syntology's identity for that exact code. It is printed in full (16 hexadecimal characters for every sample in this build), and get_code(code_sha256="…") takes it exactly as printed. An agent calls it on Syntology's MCP service (how to connect); click an identity to copy that call.

Repository labels, per sample. official repository: The archive marks this repository official for the paper. named in the paper: The archive records that the paper mentions this repository; it is not marked official. community (archive-listed): In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper. found in paper text by Syntology: Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted. community: Not in the archive's code links for this paper; a community repository Syntology harvested. Samples are grouped by repository: a repository marked official first, then by how directly the repository is tied to the paper, then in the archive's order; never by how many ran. Licence labels name the repository's licence as recorded at harvest. “Pointer only” means Syntology does not serve that copy's text, for one of six reasons: no licence file was found; the licence was not identified; the licence is recorded as permissive but that copy's record is not marked cleared; the licence is outside the permissive list Syntology serves text under (MIT, Apache-2.0, BSD and similar); this paper's copy was not located, so its licence is not recorded; or this paper's copy is cleared, but the stored copy of the same code that Syntology serves is not. Some licences outside that list permit redistribution, such as WTFPL, and GPL-3.0 under its conditions; they are simply not on the list. Hover a licence label for the reason. File links open the file on GitHub at the default branch, which may have changed since the harvest.

A function or class name links to its page when the name occurs in the harvested code of at least 5 papers; the name is a string, not an identity.

Xiaoyu-Tao/ScholarSum found in paper text by Syntology 40 samples, 31 ran (15 instrument-failed) · 9 unverified
KGChunk src/papersum/pipeline.py ran · metamorphic tier: deterministic MIT (permissive) · 2028f8749b9ea5a3 · report fixture: returned KGChunk · level recorded 2026-09-22
KGCommunity src/papersum/pipeline.py ran · metamorphic tier: deterministic MIT (permissive) · 9442043513ec9c8c · report fixture: returned KGCommunity · level recorded 2026-09-22
KGEntity src/papersum/pipeline.py ran · metamorphic tier: deterministic MIT (permissive) · c73ef73fc2219acc · report fixture: returned KGEntity · level recorded 2026-09-22
KGRelation src/papersum/pipeline.py ran · metamorphic tier: deterministic MIT (permissive) · 789ea9b770b81afa · report fixture: returned KGRelation · level recorded 2026-09-22
ModelSettings src/papersum/pipeline.py ran · metamorphic tier: deterministic MIT (permissive) · 73fa90a0662165cb · report fixture: returned ModelSettings · level recorded 2026-09-22
OpenAICompatClient src/papersum/pipeline.py ran MIT (permissive) · fe13037e86b3229e · report fixture: returned OpenAICompatClient · level recorded 2026-09-22
PaperSumResult src/papersum/pipeline.py ran · metamorphic tier: deterministic MIT (permissive) · 051462288a4de6bf · report fixture: returned PaperSumResult · level recorded 2026-09-22
ReferenceAbstract src/papersum/pipeline.py ran · metamorphic tier: deterministic MIT (permissive) · b0124ce1aaa01894 · report fixture: returned ReferenceAbstract · level recorded 2026-09-22
ReferenceCorpusRecord src/papersum/pipeline.py ran · metamorphic tier: deterministic MIT (permissive) · 4763d6d70082d3a2 · report fixture: returned ReferenceCorpusRecord · level recorded 2026-09-22
ReferenceIndexPaths src/papersum/pipeline.py ran · metamorphic tier: deterministic MIT (permissive) · 8620b947dc33b942 · report fixture: returned ReferenceIndexPaths · level recorded 2026-09-22
StudentOutput src/papersum/pipeline.py ran · metamorphic tier: deterministic MIT (permissive) · ed2130a29205de92 · report fixture: returned StudentOutput · level recorded 2026-09-22
TeacherDecision src/papersum/pipeline.py ran · metamorphic tier: deterministic MIT (permissive) · 62c79fa06bc19419 · report fixture: returned TeacherDecision · level recorded 2026-09-22
_build_messages src/papersum/pipeline.py ran · our draft was wrong fingerprinted MIT (permissive) · c0533a29bc611155 · report fixture: signature prompt:str/0/str, system_prompt:str/0/str, history_messages:list/1/dict · returned list · level recorded 2026-09-22
_cosine_similarity src/papersum/pipeline.py ran · violated contract fingerprinted MIT (permissive) · 955732112dddaec2 · report fixture: signature vector:ndarray/1/float, matrix:ndarray/2/float · returned ndarray · level recorded 2026-09-22
_dedupe_preserve_order src/papersum/pipeline.py ran · our draft was wrong MIT (permissive) · 8341fc450f8ed6c7 · report fixture: signature values:list/1/str · returned list · level recorded 2026-09-22
_normalize_name src/papersum/pipeline.py ran · our draft was wrong fingerprinted MIT (permissive) · 64a52cf00420c447 · report fixture: signature name:str/0/str · returned str · level recorded 2026-09-22
_normalize_rows src/papersum/pipeline.py ran · violated contract fingerprinted MIT (permissive) · 9c9fdc4cf37fedc9 · report fixture: signature matrix:ndarray/2/float · returned ndarray · level recorded 2026-09-22
_normalize_vector src/papersum/pipeline.py ran · violated contract fingerprinted MIT (permissive) · 8ef8c2da45a71ea0 · report fixture: signature vector:ndarray/1/float · returned ndarray · level recorded 2026-09-22
_parse_dotenv_value src/papersum/pipeline.py ran · our draft was wrong fingerprinted MIT (permissive) · e859e690868e6c9d · report fixture: signature raw_value:str/0/str · returned str · level recorded 2026-09-22
_ratio_tag src/papersum/pipeline.py ran · our draft was wrong fingerprinted MIT (permissive) · 6f2d43ed53287c21 · report fixture: signature sample_ratio:float/0/float · returned str · level recorded 2026-09-22
_safe_json_loads src/papersum/pipeline.py ran · our draft was wrong fingerprinted MIT (permissive) · 87980d14c3662951 · report fixture: signature text:str/0/str · returned dict · level recorded 2026-09-22
extract_final_answer in 14 papers src/papersum/pipeline.py ran · our draft was wrong fingerprinted MIT (permissive) · 6d46a27345390734 · report fixture: signature raw_response:str/0/str · returned str · level recorded 2026-09-22
extract_improvement_suggestions src/papersum/pipeline.py ran · our draft was wrong fingerprinted MIT (permissive) · 4f724b76ace37d5e · report fixture: signature response:str/0/str · returned str · level recorded 2026-09-22
iter_reference_mappings src/papersum/pipeline.py ran · our draft was wrong MIT (permissive) · 3aba9c10b9d46438 · report fixture: signature corpus_path:str/0/str · returned generator · level recorded 2026-09-22
load_reference_records src/papersum/pipeline.py ran · our draft was wrong MIT (permissive) · 56c191699c1f0f40 · report fixture: signature corpus_path:PosixPath/0/? · returned list · level recorded 2026-09-22
normalize_reference_text src/papersum/pipeline.py ran · our draft was wrong fingerprinted MIT (permissive) · f5c7e2e56bbb9090 · report fixture: signature text:str/0/str · returned str · level recorded 2026-09-22
parse_precision_score src/papersum/pipeline.py ran · honoured contract fingerprinted MIT (permissive) · 06d6d6889d1c5e3c · report fixture: signature response:str/0/str · returned int · level recorded 2026-09-22
sample_reference_records src/papersum/pipeline.py ran · our draft was wrong fingerprinted MIT (permissive) · 76bea012dd0cffa4 · report fixture: signature records:list/1/ReferenceCorpusRecord, sample_ratio:float/0/float, seed:int/0/int · returned list · level recorded 2026-09-22
split_indices in 5 papers src/papersum/pipeline.py ran · our draft was wrong fingerprinted MIT (permissive) · 95f3f0f1b54ce2be · report fixture: signature length:int/0/int, sample_ratio:float/0/float, seed:int/0/int · returned tuple · level recorded 2026-09-22
split_reference_records src/papersum/pipeline.py ran · our draft was wrong fingerprinted MIT (permissive) · 9758b80ff2151b82 · report fixture: signature records:list/1/ReferenceCorpusRecord, sample_ratio:float/0/float, seed:int/0/int · returned tuple · level recorded 2026-09-22
truncate_text in 9 papers src/papersum/pipeline.py ran · our draft was wrong fingerprinted MIT (permissive) · 262515fb6a504a33 · report fixture: signature text:str/0/str, max_chars:int/0/int · returned str · level recorded 2026-09-22
KGReasoningAdapter src/papersum/pipeline.py unverified MIT (permissive) · 4189e7f4dfed36de · report
LightweightKGEngine src/papersum/pipeline.py unverified MIT (permissive) · 65c0db201b0bf655 · report
PaperSumPipeline src/papersum/pipeline.py unverified MIT (permissive) · 9542d0c33f815bcf · report
PaperSumSettings src/papersum/pipeline.py unverified MIT (permissive) · bc600419c220c6e7 · report
ReferenceRetriever src/papersum/pipeline.py unverified MIT (permissive) · c26c621b2333e90c · report
StudentModule src/papersum/pipeline.py unverified MIT (permissive) · 55476bf50648d4dc · report
TeacherModule src/papersum/pipeline.py unverified MIT (permissive) · ae3af80bd1fd04d1 · report
_first_env src/papersum/pipeline.py unverified MIT (permissive) · 3ac78e7d45289d0b · report
_load_dotenv src/papersum/pipeline.py unverified MIT (permissive) · 75c1ba75133875a9 · report

Method as code Syntology

What Syntology's graph records as this paper's claimed method, lifted into code: each assembly names its bricks (functions and classes), each with its own role and its own verification level. An assembly is not a check that the code implements the method correctly or reproduces the paper's results. Levels are each brick's own verification level in Syntology's graph (level 0: no recorded run), listed brick by brick and never combined into one level; “ran” means the brick executed on a synthesized input, not that its output is correct, and the instrument figure counts failures of Syntology's instrument, not of the code. Where a model read the code against the paper's claim, each assembly's label verdict is one model's judgement, kept in this page's JSON twin; it is not calibrated, so no card prints it.

1 assembly on this paper: 1 harvested, 0 generated.

harvested Assembly for ScholarSum (graph method: ScholarSum). 40 bricks; individually on a fixture: 31 ran (of which 12 constructed an object rather than computing a result; 16 with no instrument failure: 1 honoured, 3 violated, 12 with no contract checked; 15 where Syntology's instrument failed) · 9 unverified. The assembly has not been run end-to-end.

  1. PaperSumPipeline entry · level 0 · unverified · 9542d0c33f815bcf
  2. KGReasoningAdapter model · level 0 · unverified · 4189e7f4dfed36de
  3. OpenAICompatClient helper · level 1 · ran · fe13037e86b3229e
  4. PaperSumResult helper · level 1 · ran · 051462288a4de6bf
  5. PaperSumSettings helper · level 0 · unverified · bc600419c220c6e7
  6. ReferenceRetriever data_transform · level 0 · unverified · c26c621b2333e90c
  7. StudentModule model · level 0 · unverified · 55476bf50648d4dc
  8. TeacherModule model · level 0 · unverified · ae3af80bd1fd04d1
  9. LightweightKGEngine helper · level 0 · unverified · 65c0db201b0bf655
  10. ModelSettings helper · level 1 · ran · 73fa90a0662165cb
  11. _build_messages helper · level 1 · ran · our draft was wrong · c0533a29bc611155
  12. _first_env helper · level 0 · unverified · 3ac78e7d45289d0b
  13. _load_dotenv helper · level 0 · unverified · 75c1ba75133875a9
  14. ReferenceAbstract helper · level 1 · ran · b0124ce1aaa01894
  15. ReferenceCorpusRecord helper · level 1 · ran · 4763d6d70082d3a2
  16. ReferenceIndexPaths helper · level 1 · ran · 8620b947dc33b942
  17. _normalize_rows helper · level 1 · ran · violated contract · 9c9fdc4cf37fedc9
  18. _normalize_vector helper · level 1 · ran · violated contract · 8ef8c2da45a71ea0
  19. _ratio_tag helper · level 1 · ran · our draft was wrong · 6f2d43ed53287c21
  20. load_reference_records helper · level 1 · ran · our draft was wrong · 56c191699c1f0f40
  21. normalize_reference_text helper · level 1 · ran · our draft was wrong · f5c7e2e56bbb9090
  22. sample_reference_records helper · level 1 · ran · our draft was wrong · 76bea012dd0cffa4
  23. StudentOutput helper · level 1 · ran · ed2130a29205de92
  24. extract_final_answer helper · level 1 · ran · our draft was wrong · 6d46a27345390734
  25. TeacherDecision helper · level 1 · ran · 62c79fa06bc19419
  26. extract_improvement_suggestions helper · level 1 · ran · our draft was wrong · 4f724b76ace37d5e
  27. parse_precision_score helper · level 1 · ran · honoured contract · 06d6d6889d1c5e3c
  28. truncate_text helper · level 1 · ran · our draft was wrong · 262515fb6a504a33
  29. KGChunk helper · level 1 · ran · 2028f8749b9ea5a3
  30. KGCommunity helper · level 1 · ran · 9442043513ec9c8c
  31. KGEntity helper · level 1 · ran · c73ef73fc2219acc
  32. KGRelation helper · level 1 · ran · 789ea9b770b81afa
  33. _cosine_similarity helper · level 1 · ran · violated contract · 955732112dddaec2
  34. _normalize_name helper · level 1 · ran · our draft was wrong · 64a52cf00420c447
  35. _safe_json_loads helper · level 1 · ran · our draft was wrong · 87980d14c3662951
  36. _parse_dotenv_value helper · level 1 · ran · our draft was wrong · e859e690868e6c9d
  37. iter_reference_mappings helper · level 1 · ran · our draft was wrong · 3aba9c10b9d46438
  38. split_reference_records helper · level 1 · ran · our draft was wrong · 9758b80ff2151b82
  39. _dedupe_preserve_order helper · level 1 · ran · our draft was wrong · 8341fc450f8ed6c7
  40. split_indices helper · level 1 · ran · our draft was wrong · 95f3f0f1b54ce2be

Each brick ends with its code_sha256, Syntology's identity for that exact code; click one to copy the get_code call that returns it. For agents, on Syntology's MCP service: assembly(arxiv_id="2606.18850") returns these assemblies with each brick's role and wiring (how to connect).

Results from the paper

The Papers with Code archive ends with its 2025-07-28 snapshot. This paper's arXiv identifier, 2606.18850, was issued in June 2026, after that date, so the archive has no leaderboard rows for it.

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections