Paper Citation Record · LEDGER

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference

As of 5 August 2026, this Paper Citation Record lists 27 of 27 outbound references and 3 inbound Pith citation observations for arXiv:2604.15409.

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

pith.paper-citation-record.v1
2604.15409 v1

Coverage vector

measured 27 of 27 reference resolution

Typed states for the displayed outbound observations.

Source: paper_references, paper_reference_links, observed 2026-05-10T11:23:37.593409Z

measured 30 of 30 standing notices

One-hop event checks from named stored sources.

Source: scholarly_work_events, retraction_status_cache, observed 2026-08-05T06:32:48.257954+00:00

measured 3 of 3 inbound itemization

Pith citing papers itemized under the disclosed page cap.

Source: paper_references, paper_reference_links, observed 2026-07-31T17:55:53.036918Z

measured 1 of 1 external citation measurements

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

Source: pith, observed 2026-08-05T02:28:24.338817Z

Reference resolution

27 of 27 outbound references displayed

verified exact14
verified fuzzy12
unresolved1
parse uncertain0
malformed identifier0
metadata mismatch0

External citation measurements

0
pith, observed 2026-08-05T02:28:24.338817Z

Outbound references

Observation 661c529c-fd51-431a-8b93-d23ab34eaf2c · outbound

This paper cites GPT-4 Technical Report.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference GPT-4 Technical Report

Reference 1

Resolution: verified exact; local_arxiv, observed 2026-05-10T11:25:18.692459Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation 86cffae4-8640-4fae-a6fa-e5badb453d4d · outbound

This paper cites Gqa: Training generalized multi-query transformer models from multi-head checkpoints.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Gqa: Training generalized multi-query transformer models from multi-head checkpoints

Reference 2

Resolution: verified fuzzy; raw_fallback, observed 2026-05-19T14:13:06.897271Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation 9e27fbea-6f52-44fb-a712-2bbebbbe689a · outbound

This paper cites Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Bender, Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell

Reference 3

Resolution: verified fuzzy; raw_fallback, observed 2026-05-19T14:13:06.906022Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation 8b2ec6b7-3916-403c-ba46-5325e76590ab · outbound

This paper cites How is ChatGPT's behavior changing over time?.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference How is ChatGPT's behavior changing over time?

Reference 4

Resolution: verified exact; arxiv_id, observed 2026-05-10T11:25:18.687121Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation 7dad6034-eb72-4b31-a1e3-d091bf4b1197 · outbound

This paper cites Training Verifiers to Solve Math Word Problems.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Training Verifiers to Solve Math Word Problems

Reference 5

Resolution: verified exact; local_arxiv, observed 2026-05-10T11:25:18.697465Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation 0970427e-8d47-481a-ba30-66e3a84cbe4c · outbound

This paper cites FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning

Reference 6

Resolution: verified exact; arxiv_id, observed 2026-05-11T02:39:45.042453Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation d3b2bc44-f23a-4244-be0b-5a86940275ce · outbound

This paper cites Flashattention: Fast and memory- efficient exact attention with io-awareness.Advances in neural information processing systems, 35:16344–16359.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Flashattention: Fast and memory- efficient exact attention with io-awareness.Advances in neural information processing systems, 35:16344–16359

Reference 7

Resolution: verified fuzzy; raw_fallback, observed 2026-05-19T14:13:06.908264Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation ce109177-3cdb-47ef-a655-108d6fd37bc8 · outbound

This paper cites an unresolved cited work.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Unresolved cited work

Reference 8

Resolution: unresolved; raw_fallback, observed 2026-05-19T14:13:06.903757Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation 017bad95-7e2f-40c6-a567-c4236440a8bf · outbound

This paper cites GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers

Reference 9

Resolution: verified exact; arxiv_id, observed 2026-05-10T17:18:35.295691Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation fbda724b-59c5-4090-a547-4493408c19e9 · outbound

This paper cites What every computer scientist should know about floating-point arithmetic.ACM computing surveys (CSUR), 23(1):5–48.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference What every computer scientist should know about floating-point arithmetic.ACM computing surveys (CSUR), 23(1):5–48

Reference 10

Resolution: verified fuzzy; raw_fallback, observed 2026-05-19T14:13:06.917797Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation 02fda7fb-c696-4069-b16d-3dc9013c2fb2 · outbound

This paper cites Deep learning with limited numerical precision.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Deep learning with limited numerical precision

Reference 11

Resolution: verified fuzzy; raw_fallback, observed 2026-05-19T14:13:06.912874Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation 6e93697d-9e3a-429c-a104-9a4b6790e8f8 · outbound

This paper cites Mistral 7B.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Mistral 7B

Reference 12

Resolution: verified exact; local_arxiv, observed 2026-05-10T11:25:18.645444Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation a4690602-d478-4ff3-aedc-4485288789f2 · outbound

This paper cites Efficient memory management for large language model serving with pagedattention.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Efficient memory management for large language model serving with pagedattention

Reference 13

Resolution: verified fuzzy; raw_fallback, observed 2026-05-19T14:13:06.919953Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation 15033d06-6573-41a5-835a-196eeb6db007 · outbound

This paper cites Awq: Activation-aware weight quantization for on-device llm compression and acceleration.Proceedings of machine learning and systems, 6:87–100.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Awq: Activation-aware weight quantization for on-device llm compression and acceleration.Proceedings of machine learning and systems, 6:87–100

Reference 14

Resolution: verified fuzzy; raw_fallback, observed 2026-05-19T14:13:06.914932Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation dd3f4607-b06e-48a3-a26d-f66d1d83084b · outbound

This paper cites Locating and editing factual associ- ations in gpt.Advances in neural information processing systems, 35:17359–17372.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Locating and editing factual associ- ations in gpt.Advances in neural information processing systems, 35:17359–17372

Reference 15

Resolution: verified fuzzy; raw_fallback, observed 2026-05-19T14:13:06.910934Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation 18af856f-b8c6-44d9-a100-00d1ff2a4499 · outbound

This paper cites Mixed Precision Training.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Mixed Precision Training

Reference 16

Resolution: verified exact; arxiv_id, observed 2026-05-12T10:47:18.676998Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation 59559f08-a85e-4088-870c-f528c915224c · outbound

This paper cites In-context Learning and Induction Heads.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference In-context Learning and Induction Heads

Reference 17

Resolution: verified exact; arxiv_id, observed 2026-05-11T03:49:09.911598Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation b706a312-0f44-4202-a3bd-469335a1445a · outbound

This paper cites Fast Transformer Decoding: One Write-Head is All You Need.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Fast Transformer Decoding: One Write-Head is All You Need

Reference 18

Resolution: verified exact; arxiv_id, observed 2026-05-10T23:50:20.836519Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation 2288ca89-afad-42dd-876e-14d99694b182 · outbound

This paper cites Gemma 2: Improving Open Language Models at a Practical Size.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Gemma 2: Improving Open Language Models at a Practical Size

Reference 19

Resolution: verified exact; arxiv_id, observed 2026-05-10T12:11:16.746994Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation aeb31884-3844-4cd4-8ba1-5251fdf2e3e8 · outbound

This paper cites Llama 2: Open Foundation and Fine-Tuned Chat Models.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Llama 2: Open Foundation and Fine-Tuned Chat Models

Reference 20

Resolution: verified exact; local_arxiv, observed 2026-05-10T11:25:18.662564Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation e0d2ec3d-5e3a-4188-8d89-d00528f5b84a · outbound

This paper cites Investigating gender bias in language models using causal mediation analysis.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Investigating gender bias in language models using causal mediation analysis

Reference 21

Resolution: verified fuzzy; raw_fallback, observed 2026-05-19T14:13:06.899053Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation 0c9a623c-fb56-483c-b557-82967e02c085 · outbound

This paper cites Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small

Reference 22

Resolution: verified exact; arxiv_id, observed 2026-05-13T17:13:51.588363Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation d9c4e98a-f3ef-44e4-89ce-82fab1052b45 · outbound

This paper cites Smoothquant: Ac- curate and efficient post-training quantization for large language models.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Smoothquant: Ac- curate and efficient post-training quantization for large language models

Reference 23

Resolution: verified fuzzy; raw_fallback, observed 2026-05-19T14:13:06.901614Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation 3a421cae-c5a1-4e45-8e49-c67f0a73289f · outbound

This paper cites Efficient Streaming Language Models with Attention Sinks.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Efficient Streaming Language Models with Attention Sinks

Reference 24

Resolution: verified exact; arxiv_id, observed 2026-05-11T00:40:13.867138Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation bdaa3a2b-95ef-4581-b717-881e329a1fff · outbound

This paper cites Understanding and mitigating numerical sources of nondeterminism in LLM inference.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Understanding and mitigating numerical sources of nondeterminism in LLM inference

Reference 25

Resolution: verified fuzzy; raw_fallback, observed 2026-05-19T14:13:06.892124Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation 826f8182-ee53-4443-85be-46d1bef2a8fb · outbound

This paper cites Siren's Song in the AI Ocean: A Survey on Hallucination in Large Language Models.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Siren's Song in the AI Ocean: A Survey on Hallucination in Large Language Models

Reference 26

Resolution: verified exact; arxiv_id, observed 2026-05-12T14:21:16.592379Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Observation ce4bfa4e-385f-4c54-9b77-ca9623580caa · outbound

This paper cites Judging llm-as-a-judge with mt-bench and chatbot arena.Advances in neural information processing systems, 36:46595–46623.

The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference Judging llm-as-a-judge with mt-bench and chatbot arena.Advances in neural information processing systems, 36:46595–46623

Reference 27

Resolution: verified fuzzy; raw_fallback, observed 2026-05-19T14:13:06.896994Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.

Pith citing papers

Observation 74aa9982-f46b-4ea9-bbb7-de0fc50b3a6e · inbound

From Expert Reduction to Behavioral Divergence: Tracing Numerical State through Sparse MoE Inference cites this paper.

From Expert Reduction to Behavioral Divergence: Tracing Numerical State through Sparse MoE Inference The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference

Reference 2

Resolution: unresolved; no resolver link, observed 2026-07-31T17:55:53.036918Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation a351316b-48ea-4ddb-bcbc-df48fec099f1 · inbound

Stage-Replay Divergence Follows the KV Cache: Fixed-Prefix Precision Controls and Bidirectional Cache Transplantation cites this paper.

Stage-Replay Divergence Follows the KV Cache: Fixed-Prefix Precision Controls and Bidirectional Cache Transplantation The Illusion of Equivalence: Systematic FP16 Divergence in KV-Cached Autoregressive Inference

Reference 1

Resolution: unresolved; no resolver link, observed 2026-07-31T05:35:04.782050Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 6ce380f6-267c-4add-9d6d-5317716303c5 · inbound

Stage-Replay Divergence Follows the KV Cache: Fixed-Prefix Precision Controls and Bidirectional Cache Transplantation cites this paper.

Reference 2026

Resolution: verified exact; local_arxiv, observed 2026-07-31T05:36:28.814668Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-05T06:32:48.257954+00:00.