Jiang, Jingyan, Bondapalli, Akhilesh, Rauniyar, Krishna, Keceli, Murat, Ghadar, Yasaman, Som, Haritha Siddabathuni, Vishwanath, Venkatram, Papka, Michael E., Ge, Rong, Zheng, Huihuo · Nov 11, 2026
Jiang and coauthors present AskALCF for improving HPC user support
Markdown · JSON · Original source
Jingyan Jiang and coauthors present AskALCF, an log-analysis and improvement pipeline for high-performance computing (HPC) user support. It examines production interactions, assesses answer quality and traces failures to possible causes, including missing knowledge, retrieval, prompting, model behavior or tool execution. A two-scope sufficiency test distinguishes information absent from the documentation corpus from information that is present but not retrieved. The system routes diagnoses to repairs; a controlled harness-training process uses expert references, null edits and explicit ship, revert or decline decisions. Across 2,174 support turns, the authors report that unattributable failures fell from 41.4% to 14.2%, and that expert agreement rose from 108/150 to 123/150 turns after harness training. They characterize this as a governed form of self-evolution, in which production evidence drives targeted improvements beyond periodic knowledge-base updates. The reported measures describe the system and interactions studied; the abstract does not establish how well the approach transfers to other support settings.
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
cs.AI · Article
Source details
Original source date: Nov 11, 2026 (2026-11-11).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-13T13:47:12.683Z. First observed: 2026-11-13T07:39:49.403Z. Retrieved: 2026-11-13T13:34:08.126Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
The abstract reports results across 2,174 support turns but does not specify the comparison setup or the period covered. The reported results concern the system and support interactions described in the paper; the abstract does not report results in other support settings.
Open question
How would AskALCF’s reported results change when the system is used in other support settings?
Source-stated
The authors present AskALCF as a pipeline that analyzes production support interactions, identifies possible failure sources and routes failures to repairs.
AskALCF evaluates production interactions at a fine-grained level, identifies whether failures arise from missing knowledge, retrieval, prompting, model behavior, or tool execution, and routes each failure to an actionable repair.
Source-stated
Its sufficiency test distinguishes missing documentation from information that exists but was not retrieved.
a two-scope sufficiency test that distinguishes information missing from the documentation corpus from information present but not retrieved.
Source-stated
The authors report that, across 2,174 support turns, unattributable failures decreased from 41.4% to 14.2%.
Across 2,174 support turns, AskALCF reduced unattributable failures from 41.4% to 14.2%
Source-stated
The authors report that harness training raised expert agreement from 108/150 to 123/150 turns.
harness training improved expert agreement from 108/150 to 123/150 turns.
Source-stated
The authors characterize the approach as governed self-evolution, with production evidence guiding targeted system improvements.
These results demonstrate a governed form of self-evolution in which production evidence drives targeted system improvements.
Source and contributor context
Jiang, Jingyan, Bondapalli, Akhilesh, Rauniyar, Krishna, Keceli, Murat, Ghadar, Yasaman, Som, Haritha Siddabathuni, Vishwanath, Venkatram, Papka, Michael E., Ge, Rong, Zheng, Huihuo · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Agentic AI
AI systems that plan and carry out a series of steps toward a goal, often by using other software, rather than giving one answer.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
TIME
TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Editorial provenance
Version 1. Publication ID: a60bc2f7-17f6-4bb7-86a4-0da14fd439d0.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-13T13:47:12.683Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 3126d8c7dfbfc0e4922f727854d2e24bb8c49d7055a5a40c466c79693a07f59c. Source text hash: 7a35514b16e219fc7cb4e88839f24bd01be33a4459acfec956a4229ae9d73d62.
Revision and source-check history
Version 1 · 2026-11-13T13:47:12.683Z
Jiang and coauthors present AskALCF for improving HPC user support
Jingyan Jiang and coauthors present AskALCF, an log-analysis and improvement pipeline for high-performance computing (HPC) user support. It examines production interactions, assesses answer quality and traces failures to possible causes, including missing knowledge, retrieval, prompting, model behavior or tool execution. A two-scope sufficiency test distinguishes information absent from the documentation corpus from information that is present but not retrieved. The system routes diagnoses to repairs; a controlled harness-training process uses expert references, null edits and explicit ship, revert or decline decisions. Across 2,174 support turns, the authors report that unattributable failures fell from 41.4% to 14.2%, and that expert agreement rose from 108/150 to 123/150 turns after harness training. They characterize this as a governed form of self-evolution, in which production evidence drives targeted improvements beyond periodic knowledge-base updates. The reported measures describe the system and interactions studied; the abstract does not establish how well the approach transfers to other support settings.
Original source
cs.AI · Article
Source details
Original source date: Nov 11, 2026 (2026-11-11).
Date basis: Date basis not provided.
Added to Escape Meridian: 2026-11-13T13:47:12.683Z. First observed: 2026-11-13T07:39:49.403Z. Retrieved: 2026-11-13T13:34:08.126Z.
Limits
The abstract reports results across 2,174 support turns but does not specify the comparison setup or the period covered. The reported results concern the system and support interactions described in the paper; the abstract does not report results in other support settings.
Open question
How would AskALCF’s reported results change when the system is used in other support settings?
Source-stated
The authors present AskALCF as a pipeline that analyzes production support interactions, identifies possible failure sources and routes failures to repairs.
AskALCF evaluates production interactions at a fine-grained level, identifies whether failures arise from missing knowledge, retrieval, prompting, model behavior, or tool execution, and routes each failure to an actionable repair.
Source-stated
Its sufficiency test distinguishes missing documentation from information that exists but was not retrieved.
a two-scope sufficiency test that distinguishes information missing from the documentation corpus from information present but not retrieved.
Source-stated
The authors report that, across 2,174 support turns, unattributable failures decreased from 41.4% to 14.2%.
Across 2,174 support turns, AskALCF reduced unattributable failures from 41.4% to 14.2%
Source-stated
The authors report that harness training raised expert agreement from 108/150 to 123/150 turns.
harness training improved expert agreement from 108/150 to 123/150 turns.
Source-stated
The authors characterize the approach as governed self-evolution, with production evidence guiding targeted system improvements.
These results demonstrate a governed form of self-evolution in which production evidence drives targeted system improvements.
Source and contributor context
Jiang, Jingyan, Bondapalli, Akhilesh, Rauniyar, Krishna, Keceli, Murat, Ghadar, Yasaman, Som, Haritha Siddabathuni, Vishwanath, Venkatram, Papka, Michael E., Ge, Rong, Zheng, Huihuo · Article author
These are credits stated by the source; they do not establish a profile match or authorship of every passage.
Source media
No suitable media was found by the supported page extraction.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Agentic AI
AI systems that plan and carry out a series of steps toward a goal, often by using other software, rather than giving one answer.
arXiv
arXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
TIME
TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Editorial provenance
Version 1. Publication ID: a60bc2f7-17f6-4bb7-86a4-0da14fd439d0.
Prepared by gpt-6-luna; grounding review by gpt-6-luna.
Source check: uncertain by anthropic/claude-opus-5.5 on 2026-11-13T13:47:12.683Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: material_development.
Observation: 3126d8c7dfbfc0e4922f727854d2e24bb8c49d7055a5a40c466c79693a07f59c. Source text hash: 7a35514b16e219fc7cb4e88839f24bd01be33a4459acfec956a4229ae9d73d62.
uncertain · 2026-11-13T13:47:12.683Z · anthropic/claude-opus-5.5
Publication: a60bc2f7-17f6-4bb7-86a4-0da14fd439d0
attribution — pass: The reading correctly credits Jingyan Jiang and coauthors with presenting AskALCF. It frames the results and the 'governed form of self-evolution' characterization as the authors' own report, not as established fact.
Authors:Jingyan Jiang, Akhilesh Bondapalli, Krishna Rauniyar, Murat Keceli
summary — uncertain: The summary is faithful, but its central results are hard for a non-specialist to read. 'Unattributable failures fell from 41.4% to 14.2%' could be misread as fewer support failures. The source describes a diagnostic metric: failures the system could not trace to a cause. 'Expert agreement' and 'harness training' are also unexplained; they refer to the diagnostic pipeline agreeing with expert judgments. Smallest repairs: gloss 'unattributable failures' as failures whose cause could not be identified. Clarify that expert agreement measures the diagnoses. Add term-library tooltips for 'agentic', 'retrieval' and 'harness training'.
AskALCF reduced unattributable failures from 41.4% to 14.2%, while harness training improved expert agreement from 108/150 to 123/150 turns.
date — pass: The original publication date of 2026-11-11 matches the arXiv v1 submission date. The later feed date was not used.
[Submitted on 11 Nov 2026]
support — pass: Each claim is directly backed by the abstract. The limitations accurately note that the abstract gives no comparison setup and no results outside the studied setting. The transferability question follows from the evidence without presupposing an answer. The reading does not overstate the results as real-world reliability gains.
These results demonstrate a governed form of self-evolution in which production evidence drives targeted system improvements.
media — pass: The selected media is the verified direct arXiv abstract page for this paper, which is the most relevant way to access the source.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
AI systems that plan and carry out a series of steps toward a goal, often by using other software, rather than giving one answer.
Definition referencearXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivTIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Explore TIMEAI systems that plan and carry out a series of steps toward a goal, often by using other software, rather than giving one answer.
Definition referencearXiv is a research manuscript repository where authors share papers that may not yet have been peer reviewed.
Explore arXivTIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Explore TIME