Paper Citation Record · LEDGER

Multi-Head Attention Residuals

As of 8 August 2026, this Paper Citation Record lists 26 of 26 outbound references and 0 inbound Pith citation observations for arXiv:2607.27230.

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

pith.paper-citation-record.v1
2607.27230 v2

Coverage vector

measured 26 of 26 reference resolution

Typed states for the displayed outbound observations.

Source: paper_references, paper_reference_links, observed 2026-08-04T01:39:49.827447Z

measured 26 of 26 standing notices

One-hop event checks from named stored sources.

Source: scholarly_work_events, retraction_status_cache, observed 2026-08-08T06:32:00.761636+00:00

measured 0 of 0 inbound itemization

Pith citing papers itemized under the disclosed page cap.

Source: paper_references, paper_reference_links

measured 0 of 1 external citation measurements

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

Source: cited_works

Reference resolution

26 of 26 outbound references displayed

verified exact0
verified fuzzy0
unresolved25
parse uncertain0
malformed identifier1
metadata mismatch0

External citation measurements

No source-named external measurement is stored.

Outbound references

Observation 728e9bfb-469e-4fb0-a36a-4ab7e731f525 · outbound

This paper cites Training Verifiers to Solve Math Word Problems.

Multi-Head Attention Residuals Training Verifiers to Solve Math Word Problems

Reference 3

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:47.618608Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 5f87f737-1654-432c-b6d6-123c36b929e5 · outbound

This paper cites This is the loss-level counterpart of the width-isolated disagreement in Table 6 (which rises 0.235→0.281 over the same widening).

Multi-Head Attention Residuals This is the loss-level counterpart of the width-isolated disagreement in Table 6 (which rises 0.235→0.281 over the same widening)

Reference 4

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:49.569360Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 3f3ca8a2-f101-455b-bb99-d121b105a5cf · outbound

This paper cites mHC: Manifold-Constrained Hyper-Connections.

Multi-Head Attention Residuals mHC: Manifold-Constrained Hyper-Connections

Reference 5

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:47.758808Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 1c3ef572-3e4e-4748-ac27-05cfea4c7845 · outbound

This paper cites This control is a self-contained 5×10 −4 comparison on the web corpus, so its d512 deltas differ slightly from Table 10’s tuned-rate (1×10 −3) numbers.

Multi-Head Attention Residuals This control is a self-contained 5×10 −4 comparison on the web corpus, so its d512 deltas differ slightly from Table 10’s tuned-rate (1×10 −3) numbers

Reference 10

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:49.451208Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 066941b7-06c5-4da0-b5ea-677bcaf10f87 · outbound

This paper cites Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free.

Multi-Head Attention Residuals Gated Attention for Large Language Models: Non-linearity, Sparsity, and Attention-Sink-Free

Reference 11

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:48.142895Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 11a3f45a-48b3-48c5-96c1-dc835482131c · outbound

This paper cites Mixture-of-Depths: Dynamically allocating compute in transformer-based language models.

Multi-Head Attention Residuals Mixture-of-Depths: Dynamically allocating compute in transformer-based language models

Reference 13

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:48.300907Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 3b4514cd-824d-47b6-a1ff-8b93512acde5 · outbound

This paper cites DeepNet: Scaling Transformers to 1,000 Layers.

Multi-Head Attention Residuals DeepNet: Scaling Transformers to 1,000 Layers

Reference 15

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:48.498444Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 1b4b545e-1df6-442e-887d-8467c30f4459 · outbound

This paper cites Deep Delta Learning.

Multi-Head Attention Residuals Deep Delta Learning

Reference 16

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:48.637581Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 2440badd-7c79-4ad8-a11c-bebdd9b04be3 · outbound

This paper cites an unresolved cited work.

Multi-Head Attention Residuals Unresolved cited work

Reference 17

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:48.762582Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 1f04883f-7124-4e14-b24a-b76a06508a4f · outbound

This paper cites DenseNets [Huang et al., 2017] instead concatenate all previous feature maps, giving each layer direct access to every earlier one.

Multi-Head Attention Residuals DenseNets [Huang et al., 2017] instead concatenate all previous feature maps, giving each layer direct access to every earlier one

Reference 18

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:48.845383Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 57b1671f-e634-491e-829c-25da364d1fca · outbound

This paper cites an unresolved cited work.

Multi-Head Attention Residuals Unresolved cited work

Reference 19

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:48.951279Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 3de68183-1647-46f1-a52b-0fc0f9eb984e · outbound

This paper cites ""Single-head depth routing = attention residuals (Kimi 2025): 3one shared query, one softmax over depth, read by all D coords. 4Equals the MHAR route at H=1.

Multi-Head Attention Residuals ""Single-head depth routing = attention residuals (Kimi 2025): 3one shared query, one softmax over depth, read by all D coords. 4Equals the MHAR route at H=1

Reference 20

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:49.073527Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation facae91e-418f-4e50-97d6-62e00a0dc625 · outbound

This paper cites an unresolved cited work.

Multi-Head Attention Residuals Unresolved cited work

Reference 21

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:49.154723Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation c49e995e-a5b1-42dc-8414-8c458585ca37 · outbound

This paper cites 1def route_bwd(V, W, dout, dV, dq_part, dg_part):# one program per pos.

Multi-Head Attention Residuals 1def route_bwd(V, W, dout, dV, dq_part, dg_part):# one program per pos

Reference 22

Resolution: malformed identifier; no resolver link, observed 2026-08-04T01:39:49.279598Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation acec9d32-567d-49a0-842f-c2ec04ea46e2 · outbound

This paper cites The two runs are identical except for the routing mechanism (same node, software, data order, and global batch).

Multi-Head Attention Residuals The two runs are identical except for the routing mechanism (same node, software, data order, and global batch)

Reference 25

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:49.745960Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation c97f8e7b-24f5-447a-aa26-f057f1ce44bc · outbound

This paper cites Documents are streamed from the corpus shards, joined with the end-of-text token (id 151,645), and packed into contiguous sequences of length T with no padding.

Multi-Head Attention Residuals Documents are streamed from the corpus shards, joined with the end-of-text token (id 151,645), and packed into contiguous sequences of length T with no padding

Reference 26

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:49.827447Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation a1f35164-2779-4fc8-84ae-617baf6f448c · outbound

This paper cites Nemotron-CC: Transforming Common Crawl into a Refined Long-Horizon Pretraining Dataset.

Multi-Head Attention Residuals Nemotron-CC: Transforming Common Crawl into a Refined Long-Horizon Pretraining Dataset

Reference 2015

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:48.386899Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 5b4d62da-a1ce-4e72-9d89-e3afa7859275 · outbound

This paper cites The FineWeb Datasets: Decanting the Web for the Finest Text Data at Scale.

Multi-Head Attention Residuals The FineWeb Datasets: Decanting the Web for the Finest Text Data at Scale

Reference 2016

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:48.074782Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation bd5aa1d8-db5a-42db-a10d-2648fef6ea78 · outbound

This paper cites Attention Residuals.

Multi-Head Attention Residuals Attention Residuals

Reference 2017

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:47.872763Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 7b329004-10d6-4462-83c2-3c2d4c8f0fc0 · outbound

This paper cites Delta Attention Residuals.

Multi-Head Attention Residuals Delta Attention Residuals

Reference 2019

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:47.959950Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 1879db55-f1bc-43f6-a35b-bbed25a828e3 · outbound

This paper cites Evaluating Large Language Models Trained on Code.

Multi-Head Attention Residuals Evaluating Large Language Models Trained on Code

Reference 2021

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:47.559252Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 4cc59b98-1840-4561-9278-3d5203a95c98 · outbound

This paper cites AI capabilities can be significantly improved without expensive retraining.

Multi-Head Attention Residuals AI capabilities can be significantly improved without expensive retraining

Reference 2022

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:47.687200Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation a434357d-c39c-4d8e-8c1c-98c850430191 · outbound

This paper cites Program Synthesis with Large Language Models.

Multi-Head Attention Residuals Program Synthesis with Large Language Models

Reference 2023

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:47.523424Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 480fa5e1-fb6a-4b47-b397-75347d8550e0 · outbound

This paper cites Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun.

Multi-Head Attention Residuals Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun

Reference 2024

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:47.822580Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation cfb25400-5397-4712-bb6d-1b422a71d696 · outbound

This paper cites Qwen3 Technical Report.

Multi-Head Attention Residuals Qwen3 Technical Report

Reference 2025

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:48.221064Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 88939309-ecc3-4dce-96b2-19a208537475 · outbound

This paper cites DenseFormer: Enhancing Information Flow in Transformers via Depth Weighted Averaging.

Multi-Head Attention Residuals DenseFormer: Enhancing Information Flow in Transformers via Depth Weighted Averaging

Reference 2026

Resolution: unresolved; no resolver link, observed 2026-08-04T01:39:48.022554Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Pith citing papers

No inbound Pith citation observations are available.