Pith. sign in

Paper Citation Record · LEDGER

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization

As of 17 August 2026, this Paper Citation Record lists 44 of 44 outbound references and 2 inbound Pith citation observations for arXiv:2603.23566.

A citation records a reference. It does not transfer a finding from one paper to another.

pith.paper-citation-record.v1
2603.23566 v2

Coverage vector

measured 44 of 44 reference resolution

Typed states for the displayed outbound observations.

Source: paper_references, paper_reference_links, observed 2026-05-21T10:16:15.305706Z

measured 46 of 46 standing notices

One-hop event checks from named stored sources.

Source: scholarly_work_events, retraction_status_cache, observed 2026-08-17T06:30:58.91139+00:00

measured 2 of 2 inbound itemization

Pith citing papers itemized under the disclosed page cap.

Source: paper_references, paper_reference_links, observed 2026-08-02T08:11:51.592817Z

measured 0 of 1 external citation measurements

A source-named dated measurement, never combined with another source.

Source: cited_works

Reference resolution

44 of 44 outbound references displayed

  • verified exact21
  • verified fuzzy23
  • unresolved0
  • parse uncertain0
  • malformed identifier0
  • metadata mismatch0

External citation measurements

No source-named external measurement is stored.

Outbound references

Observation e9d11abd-f7da-4dd9-b7df-52ae006db54d · outbound

This paper cites NeutronAscend: Op- timizing GNN training with Ascend AI processors.ACM Transactions on Architecture and Code Optimization, 22(4):1–26.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization NeutronAscend: Op- timizing GNN training with Ascend AI processors.ACM Transactions on Architecture and Code Optimization, 22(4):1–26

Reference 1

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.703695Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:59fc3b6d781c80fcdcc5deaab8ee7173236f6cf1729a807712a9c9f5d23b7d1d

Observation 8c322bb4-a551-431a-91c3-0f9d0beb48d0 · outbound

This paper cites GPU Kernel Scientist: An LLM-Driven Framework for Iterative Kernel Optimization.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization GPU Kernel Scientist: An LLM-Driven Framework for Iterative Kernel Optimization

Reference 2

Resolution
verified exact
arxiv_id, observed 2026-05-21T10:20:00.050268Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:8090724dce17b11605792c6cba0c28efb46332204002a007e31fb485f2269293

Observation 047fdc06-5d17-4948-9ed0-594b51feed50 · outbound

This paper cites Tiramisu: A polyhedral compiler for expressing fast and portable code.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization Tiramisu: A polyhedral compiler for expressing fast and portable code

Reference 3

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.712101Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:7f564d1d1370fc65f90339d515708a6e393d56c704c2c6182a9a3a8091f4264f

Observation 9fba3d5f-733d-4ee9-99d1-7f48f8b58973 · outbound

This paper cites Kevin: Multi-Turn RL for Generating CUDA Kernels.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization Kevin: Multi-Turn RL for Generating CUDA Kernels

Reference 4

Resolution
verified exact
arxiv_id, observed 2026-05-21T10:20:00.045508Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:fcfcbd919c10ec9d0f55f3cbb3cab5bcd27bcbd82f29fff504769422996853d9

Observation db46928d-91c5-44dc-8651-bdeabf10f8f7 · outbound

This paper cites A practical automatic polyhedral parallelizer and locality optimizer.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization A practical automatic polyhedral parallelizer and locality optimizer

Reference 5

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.720885Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:9b179b80f6f16ffe0ad7a41239892cc203b27fc9f29852ad4f52a6fe4c7c94c5

Observation 06f1982f-eb29-4f84-90d2-fde9dda89c7e · outbound

This paper cites AscendKernelGen: A Systematic Study of LLM-Based Kernel Generation for Neural Processing Units.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization AscendKernelGen: A Systematic Study of LLM-Based Kernel Generation for Neural Processing Units

Reference 6

Resolution
verified exact
local_arxiv, observed 2026-05-21T10:20:00.092462Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:1f3408848ffc26dfc44412c5491666a0e5688852d35f747896d170f85bed4762

Observation 20854c5e-7ad7-4653-9553-9a03390cc6c8 · outbound

This paper cites TVM: An automated end-to-end optimizing compiler for deep learning.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization TVM: An automated end-to-end optimizing compiler for deep learning

Reference 7

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.665632Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:967ddfed74184290adc878aff04c3c1088dd4a43ea0f50a987fa289c73a7603d

Observation b043dedf-348e-4c04-8626-a1f0bc81341c · outbound

This paper cites Flashattention: Fast and memory-efficient exact attention with IO-awareness.the 36th International Conference on Neural Information Processing Systems (NeurIPS).

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization Flashattention: Fast and memory-efficient exact attention with IO-awareness.the 36th International Conference on Neural Information Processing Systems (NeurIPS)

Reference 8

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.660193Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:fb6571a742746045c0b1d20263092baf4a8120213b489c0ddd8de76c220b11b8

Observation 5659c11b-3485-4cdd-a356-42a5690686a2 · outbound

This paper cites STARK: Strategic team of agents for refining kernels.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization STARK: Strategic team of agents for refining kernels

Reference 9

Resolution
verified exact
arxiv_id, observed 2026-05-21T10:20:00.082652Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:5496d193e87cd3dc59fcb8f7f4411c6e6bf492f18d27b3a39421f7d0c27a9093

Observation d96cc372-0955-466d-bc40-6afb8d792f0e · outbound

This paper cites EvoEngineer: Mastering automated CUDA kernel code evolution with large language models.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization EvoEngineer: Mastering automated CUDA kernel code evolution with large language models

Reference 10

Resolution
verified exact
arxiv_id, observed 2026-05-21T10:20:00.059737Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:1a297ede3204b15297248be3ddd2365ad75f49ca548096f119fdc320752e480b

Observation 6213f072-17ac-4a8c-845a-9973aa37b9a0 · outbound

This paper cites PRAGMA: A profiling-reasoned multi-agent framework for automatic kernel optimization.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization PRAGMA: A profiling-reasoned multi-agent framework for automatic kernel optimization

Reference 11

Resolution
verified exact
arxiv_id, observed 2026-05-21T10:20:00.078041Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:43dd66de0756a97e315ca2c7aa1c1a99670644dc1dd1ef13ba82f7b8b4234de5

Observation f3545ae0-6a5e-41b9-bcad-1063adea01df · outbound

This paper cites Tritonforge: Profiling-guided framework for automated triton kernel optimization.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization Tritonforge: Profiling-guided framework for automated triton kernel optimization

Reference 12

Resolution
verified exact
arxiv_id, observed 2026-05-21T10:20:00.032073Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:4668978f4fc6f1e49aeb60aa32a128533c6ece3f8993c1a9ad650439fcc298fd

Observation b2eb99bf-3937-4005-bf1f-38b0638c17e4 · outbound

This paper cites TritonBench: Benchmarking large language model capabilities for generating Triton operators.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization TritonBench: Benchmarking large language model capabilities for generating Triton operators

Reference 13

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.714972Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:1f004a500736abad826fc57d603b1735391230f9596c3e88602a2f45df94f921

Observation 0f3ded8b-6200-41b0-bd62-c41c4e3c7ea5 · outbound

This paper cites The deep learning compiler: A comprehensive survey.IEEE Transactions on Parallel and Distributed Systems, 32(3):708–727.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization The deep learning compiler: A comprehensive survey.IEEE Transactions on Parallel and Distributed Systems, 32(3):708–727

Reference 14

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.718023Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:c5866dadb8e7f3efa426717d099c6d09df6b4b3bbe98a116642a55dee231ff29

Observation 2d873a39-a829-4847-aaf2-9e3fae31f37e · outbound

This paper cites AutoTriton: Automatic Triton Programming with Reinforcement Learning in LLMs.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization AutoTriton: Automatic Triton Programming with Reinforcement Learning in LLMs

Reference 15

Resolution
verified exact
arxiv_id, observed 2026-05-21T10:20:00.064098Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:58d16aff120524585091c6fe8a84ff5858410387ba694b2814cbe8e4c4c62897

Observation 976771a4-57fc-4df7-b6dc-e5c030a340bb · outbound

This paper cites StitchCUDA: An Automated Multi-Agents End-to-End GPU Programing Framework with Rubric-based Agentic Reinforcement Learning.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization StitchCUDA: An Automated Multi-Agents End-to-End GPU Programing Framework with Rubric-based Agentic Reinforcement Learning

Reference 16

Resolution
verified exact
arxiv_id, observed 2026-08-11T03:24:38.516909Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:7accc3dd45723309e73de2b86d552ae4b795b247296713e60cbce713d774a459

Observation f0eac67c-39fc-4253-929c-82acc619deca · outbound

This paper cites CUDA-L1: Improving CUDA Optimization via Contrastive Reinforcement Learning.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization CUDA-L1: Improving CUDA Optimization via Contrastive Reinforcement Learning

Reference 17

Resolution
verified exact
arxiv_id, observed 2026-07-14T02:20:17.802860Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:5c8736f3d3a48bf7e252ccfdc9edd01f09513fa16ff5f02b921018a5e39f7260

Observation c63b94c3-8612-446c-8794-15ff1b6145bb · outbound

This paper cites KernelEvolve: Scaling Agentic Kernel Coding for Heterogeneous AI Accelerators at Meta.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization KernelEvolve: Scaling Agentic Kernel Coding for Heterogeneous AI Accelerators at Meta

Reference 18

Resolution
verified exact
arxiv_id, observed 2026-07-08T02:18:09.597449Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:51bfe2e4983078ef7644275750e0570453fe6800d1b179571666273d45f5e75b

Observation 2af18a5b-006b-4bb0-ab43-843124b2f989 · outbound

This paper cites Accelerating sparse matrix-matrix multiplication with the Ascend AI core.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization Accelerating sparse matrix-matrix multiplication with the Ascend AI core

Reference 19

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.706481Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:0e8ecdfd09e962f6663352bb4734d3a12d1c9f5be5654c540ae5bebc75e94ca5

Observation 154d02c0-c315-4c6a-a0e5-117bb5dc285b · outbound

This paper cites KernelBench: Can LLMs Write Efficient GPU Kernels?.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization KernelBench: Can LLMs Write Efficient GPU Kernels?

Reference 20

Resolution
verified exact
local_arxiv, observed 2026-05-21T10:20:00.068223Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:91e2a6defba0d8e7c52f20c09ee4d64de90ab3c469ecc022724fbb640ef7ba8b

Observation 134ecb07-f631-49fa-a30b-5e18cd51663a · outbound

This paper cites Halide: A language and compiler for optimizing parallelism, locality, and recomputation in image processing pipelines.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization Halide: A language and compiler for optimizing parallelism, locality, and recomputation in image processing pipelines

Reference 21

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.709212Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:2dbc59088042ca4adbf3568b16617e38744f3e79bd605216e1ffeb37363b600a

Observation eaec2863-bbd7-44e4-9fe2-3212a03f0968 · outbound

This paper cites Seed-Coder: Let the Code Model Curate Data for Itself.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization Seed-Coder: Let the Code Model Curate Data for Itself

Reference 22

Resolution
verified exact
arxiv_id, observed 2026-05-21T10:20:00.087467Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:50f12b895139db62ccf3d7282dde4facffb22182f96c7762445bb4c96007efb0

Observation 0eec9851-a374-4168-af5c-0e81ddc376b3 · outbound

This paper cites FlashAttention-3: Fast and accurate attention with asynchrony and low-precision.The 38th Conference on Neural Information Processing Systems (NeurIPS).

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization FlashAttention-3: Fast and accurate attention with asynchrony and low-precision.The 38th Conference on Neural Information Processing Systems (NeurIPS)

Reference 23

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.697665Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:678a66ab749cca9052b6c5e65013ea99df73c39cd3e93bb78218fda68cc7663e

Observation 74e09198-7c17-4e3e-8500-6bc869f87d39 · outbound

This paper cites OpenEvolve: An open-source evolutionary coding agent.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization OpenEvolve: An open-source evolutionary coding agent

Reference 24

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.700709Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:13ab4e9f4dac28481d58dd521b4ef4881d033ceafbb197a537394054520a72c5

Observation e89d0978-6698-4d07-a706-547c71bd5372 · outbound

This paper cites CUDA-L2: Surpassing cuBLAS Performance for Matrix Multiplication through Reinforcement Learning.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization CUDA-L2: Surpassing cuBLAS Performance for Matrix Multiplication through Reinforcement Learning

Reference 25

Resolution
verified exact
arxiv_id, observed 2026-07-14T02:20:20.617597Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:ccd1d389abe16c781634ae709480b7a52e52f86d5a1c846afc6b8ec601115721

Observation f69290e5-1204-4dec-ba99-2255c88449a7 · outbound

This paper cites Triton: An intermediate language and compiler for tiled neural network computations.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization Triton: An intermediate language and compiler for tiled neural network computations

Reference 26

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.694673Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:da73ddb097af7ad5b89da330149c377a91b6de284b1f294078abcebed4ad1bed

Observation d0d06f4e-cfc5-4d9d-b013-6a62fa2770f5 · outbound

This paper cites Geak: Introducing Triton Kernel AI Agent & Evaluation Benchmarks.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization Geak: Introducing Triton Kernel AI Agent & Evaluation Benchmarks

Reference 27

Resolution
verified exact
arxiv_id, observed 2026-05-21T10:20:00.009036Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:27561d2d5aed2145ea9653a9a1412369d4de190be5797b8006e2742ab1543f55

Observation dce00cc6-6238-47aa-8e50-d70f4497d759 · outbound

This paper cites TileLang: A Composable Tiled Programming Model for AI Systems.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization TileLang: A Composable Tiled Programming Model for AI Systems

Reference 28

Resolution
verified exact
arxiv_id, observed 2026-05-21T10:20:00.022432Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:aac288946bf80f06bc29c386d523a19b7b8fd44a9612cf2795b9507e697977cb

Observation 2a607446-d0ad-4a70-b010-b1193257e533 · outbound

This paper cites Astra: A multi-agent system for GPU kernel performance optimization.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization Astra: A multi-agent system for GPU kernel performance optimization

Reference 29

Resolution
verified exact
arxiv_id, observed 2026-05-21T10:20:00.004601Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:6ac46feda80428959d7c24565653f0e5361196bdb34a98dd58eba166567319e3

Observation f84cb9ff-3ec1-4038-93da-431015150c07 · outbound

This paper cites MultiKernelBench: A Multi-Platform Benchmark for Kernel Generation.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization MultiKernelBench: A Multi-Platform Benchmark for Kernel Generation

Reference 30

Resolution
verified exact
arxiv_id, observed 2026-05-21T10:20:00.013601Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:198fdc939daa1f28603e911dd0cf3757869da2d4d253de35463856c87bb3edd0

Observation e79dd4e9-3ec2-4d87-ad1d-c1919dab66ff · outbound

This paper cites TritonRL: Training LLMs to think and code triton without cheating.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization TritonRL: Training LLMs to think and code triton without cheating

Reference 31

Resolution
verified exact
arxiv_id, observed 2026-05-21T10:20:00.054808Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:d68b372f6c8ef16f948d44e3e80b87e23037d5c876c903e3e5200d75d0db1788

Observation a78a0716-0ba4-48b7-b45b-6b414edfc4c0 · outbound

This paper cites Mirage: A multi-level superoptimizer for tensor programs.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization Mirage: A multi-level superoptimizer for tensor programs

Reference 32

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.691753Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:9ce90bc2d436ef220adef2889bc9bcadb10edb21a49fc86f5470251433e5d251

Observation 4850ea38-86d4-4f15-bc3c-9444205126c2 · outbound

This paper cites TLP: A deep learning- based cost model for tensor program tuning.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization TLP: A deep learning- based cost model for tensor program tuning

Reference 33

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.688908Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:3bdfa470414ef7bc4ee75dab231f13cca745a1c3d48f738d81f546119f12db89

Observation bb09a0fe-b3ac-4ab8-bbbb-7a615caf0ca5 · outbound

This paper cites ReWiND: Language-guided rewards teach robot policies without new demonstrations.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization ReWiND: Language-guided rewards teach robot policies without new demonstrations

Reference 34

Resolution
verified exact
arxiv_id, observed 2026-05-21T10:20:00.073093Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:291d72c6f756ec81fdeeff05b1e437d0a58c9e076be83b91d812b472dffb5248

Observation d3561414-f557-4352-8140-ae390af48fd3 · outbound

This paper cites Cudaforge: An agent framework with hardware feedback for cuda kernel optimization.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization Cudaforge: An agent framework with hardware feedback for cuda kernel optimization

Reference 35

Resolution
verified exact
arxiv_id, observed 2026-05-21T10:20:00.040818Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:7054594051805fd00c4d00ccc2a88d4620f311c4a462c8f0532707f169340479

Observation 7f268062-034b-4a89-8e67-68ef8d637304 · outbound

This paper cites AKG: Automatic kernel generation for neural processing units using polyhedral transformations.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization AKG: Automatic kernel generation for neural processing units using polyhedral transformations

Reference 36

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.723819Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:ea8b5efdaeca632c10b97854859cc1620410366ef25e0998372e325136307bfd

Observation 9c91daf4-c080-4538-8629-25c160cd54a0 · outbound

This paper cites Ansor: Generating high-performance tensor programs for deep learning.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization Ansor: Generating high-performance tensor programs for deep learning

Reference 37

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.677517Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:8d2db49f1a80c191d2e4ceca2af3279418335ae62a73d15f9f88f0841b8eb3fb

Observation 0465cf72-6a47-4440-90bd-e7c51ad9243f · outbound

This paper cites TenSet: A large-scale program performance dataset for learned tensor compilers.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization TenSet: A large-scale program performance dataset for learned tensor compilers

Reference 38

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.686087Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:b085e2c668542894c4d0e05840bbbc0f311fac640cf78817e2b615cedb24c2bf

Observation 078ba9e3-af67-407e-a43e-2d15026dc8a4 · outbound

This paper cites Squeezing operator performance potential for the ascend architecture.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization Squeezing operator performance potential for the ascend architecture

Reference 39

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.662957Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:0cb474f08180a0cd2f181a6dee96f7b1a21270fa7e11fb47b85bc9137226fa32

Observation eb412db0-ac5b-4011-bafe-45a8b802c41c · outbound

This paper cites Accelerating model training on Ascend chips: An industrial system for profiling, analysis and optimization.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization Accelerating model training on Ascend chips: An industrial system for profiling, analysis and optimization

Reference 40

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.668616Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:1c33fa4aecda92030da93b505c6efc8720db0bc2234d75ac5446c2d4f43208e6

Observation df1539b0-d0d5-4a56-879a-058f02055268 · outbound

This paper cites increase tiling.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization increase tiling

Reference 41

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.683170Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:f25d954c9cce68bcfd2f1f9bf7520f727211fc7816d5ba223faa44a735036286

Observation 27ba7d62-daf2-4548-a355-5385aa49567a · outbound

This paper cites 7- On the Ascend AI Core, this forces a full stall across MTE1/MTE2/MTE3 and the Vector units, preventing overlap between data movement and computation.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization 7- On the Ascend AI Core, this forces a full stall across MTE1/MTE2/MTE3 and the Vector units, preventing overlap between data movement and computation

Reference 42

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.674664Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:7a2537a9994686afdefcf8b4ea22936b24078876c3ab8fb31300f9c286d7643d

Observation 8be0aec8-7e05-45a0-969a-8daad4f63b3c · outbound

This paper cites 12- Increasing it to ‘maxDataCount = 8‘ (32 bytes) improves the payload-to-overhead ratio for DMA commands.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization 12- Increasing it to ‘maxDataCount = 8‘ (32 bytes) improves the payload-to-overhead ratio for DMA commands

Reference 43

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.671938Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:bf1bb8354bee4d32a269e168ab3d0b1fce66efd0c91af5aa15e0f6d68e79666e

Observation c03ece47-65c8-4f92-8d16-c57946408282 · outbound

This paper cites 17"bottleneck.

AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization 17"bottleneck

Reference 44

Resolution
verified fuzzy
raw_fallback, observed 2026-05-21T10:20:00.680239Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-17T06:30:58.91139+00:00.

source=pdf_text observed=2026-05-21T10:16:15.305706Z digest=sha256:0014aeaa7c8aebfadd9efff646891d987e97a5b7b9ad3fbbefc1281fddf5c2da

Pith citing papers

Observation 8f70b291-131c-4bcb-af34-b4b1ede46179 · inbound

SLAI T-Rex: Full-Parameter Post-training of the DeepSeek-V4 Family on Ascend SuperPOD cites this paper.

SLAI T-Rex: Full-Parameter Post-training of the DeepSeek-V4 Family on Ascend SuperPOD AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization

Reference 46

Resolution
unresolved
no resolver link, observed 2026-08-01T10:42:40.072916Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

source=arxiv_source observed=2026-08-01T10:42:40.072916Z digest=sha256:5505961792ecf297d262e4b701f84d2b9629245013f3eb4d04f063d8395c61ec

Observation 1a3f8c0d-0bf6-4022-8c4f-df6d6299f2f4 · inbound

CANN Bench: Benchmarking Agent Generated Kernels against Real NPU and Algorithmic Limits cites this paper.

CANN Bench: Benchmarking Agent Generated Kernels against Real NPU and Algorithmic Limits AscendOptimizer: Episodic Agent for Ascend NPU Operator Optimization

Reference 13

Resolution
unresolved
no resolver link, observed 2026-08-02T08:11:51.592817Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

source=pdf_text observed=2026-08-02T08:11:51.592817Z digest=sha256:79a20bc875db0df30d3373e1ccda2ca026d91c500232c080266d15dca5be69f7