Paper Citation Record · LEDGER

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding

As of 23 August 2026, this Paper Citation Record lists 24 of 24 outbound references and 0 inbound Pith citation observations for arXiv:2606.21906.

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

pith.paper-citation-record.v1
2606.21906 v1

Coverage vector

measured 24 of 24 reference resolution

Typed states for the displayed outbound observations.

Source: paper_references, paper_reference_links, observed 2026-06-26T12:21:42.088795Z

measured 24 of 24 standing notices

One-hop event checks from named stored sources.

Source: scholarly_work_events, retraction_status_cache, observed 2026-08-22T06:32:14.747728+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

24 of 24 outbound references displayed

verified exact8
verified fuzzy0
unresolved14
parse uncertain0
malformed identifier1
metadata mismatch1

External citation measurements

No source-named external measurement is stored.

Outbound references

Observation fc86705e-cba9-471b-a09d-89369334a3e5 · outbound

This paper cites A General Language Assistant as a Laboratory for Alignment.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding A General Language Assistant as a Laboratory for Alignment

Reference 1

Resolution: verified exact; local_arxiv, observed 2026-07-04T07:59:40.105129Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.

Observation 56a2b02b-9856-476c-891e-c102a7078b3a · outbound

This paper cites Eliciting Latent Predictions from Transformers with the Tuned Lens.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding Eliciting Latent Predictions from Transformers with the Tuned Lens

Reference 2

Resolution: verified exact; local_arxiv, observed 2026-07-04T07:59:40.102550Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.

Observation 826b59a2-a2ce-4fa0-acb6-7e42761a6d72 · outbound

This paper cites Training Verifiers to Solve Math Word Problems.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding Training Verifiers to Solve Math Word Problems

Reference 3

Resolution: verified exact; local_arxiv, observed 2026-07-04T07:59:40.099944Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.

Observation 859a0e6c-20e9-48ca-b059-040568c823db · outbound

This paper cites Transformer feed-forward layers build predictions by promoting concepts in the vocabulary space.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding Transformer feed-forward layers build predictions by promoting concepts in the vocabulary space

Reference 4

Resolution: unresolved; no resolver link, observed 2026-06-26T12:21:42.088795Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation d00e2c91-06fb-4487-bfe6-95648d3f4e3a · outbound

This paper cites Dissecting recall of factual associations in auto-regressive language models.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding Dissecting recall of factual associations in auto-regressive language models

Reference 5

Resolution: unresolved; no resolver link, observed 2026-06-26T12:21:42.088795Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 952e87b1-bba6-49f9-9622-bdbbc1374099 · outbound

This paper cites How do llms use their depth?arXiv preprint arXiv:2510.18871.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding How do llms use their depth?arXiv preprint arXiv:2510.18871

Reference 6

Resolution: verified exact; arxiv_id, observed 2026-07-04T07:59:40.122455Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.

Observation c3c03b33-93ae-4c1e-9b8e-a3dc965aa4c5 · outbound

This paper cites Inference-time intervention: Eliciting truthful answers from a language model.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding Inference-time intervention: Eliciting truthful answers from a language model

Reference 7

Resolution: unresolved; no resolver link, observed 2026-06-26T12:21:42.088795Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation c3e836b1-9439-427f-9b68-4902d744ace3 · outbound

This paper cites Layer-order inversion: Rethinking latent multi-hop reasoning in large language models.arXiv preprint arXiv:2601.03542,.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding Layer-order inversion: Rethinking latent multi-hop reasoning in large language models.arXiv preprint arXiv:2601.03542,

Reference 8

Resolution: verified exact; arxiv_id, observed 2026-07-04T07:59:40.110778Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.

Observation a5ce01a8-a7d6-474a-b21e-1c559e159c14 · outbound

This paper cites Diffskip: Differential layer skipping in large language models.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding Diffskip: Differential layer skipping in large language models

Reference 9

Resolution: unresolved; no resolver link, observed 2026-06-26T12:21:42.088795Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation cfc81bef-5d02-4bb7-92fe-c6c2037968f8 · outbound

This paper cites Accessed: 2026-02-16.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding Accessed: 2026-02-16

Reference 10

Resolution: unresolved; no resolver link, observed 2026-06-26T12:21:42.088795Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation ab7cb403-a851-4411-a81c-8d902bbfba7b · outbound

This paper cites Trusting your evidence: Hallucinate less with context-aware decoding.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding Trusting your evidence: Hallucinate less with context-aware decoding

Reference 11

Resolution: unresolved; no resolver link, observed 2026-06-26T12:21:42.088795Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation abbebdb4-c20b-4a27-a55e-1885be4bd1e2 · outbound

This paper cites The diminishing returns of early-exit decoding in modern llms.arXiv preprint arXiv:2603.23701,.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding The diminishing returns of early-exit decoding in modern llms.arXiv preprint arXiv:2603.23701,

Reference 12

Resolution: verified exact; arxiv_id, observed 2026-07-04T07:59:40.116684Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.

Observation 4a0a2670-43a2-498c-8f14-7aa5062c04d8 · outbound

This paper cites Dynamic early exit in reasoning models.arXiv preprint arXiv:2504.15895, 2025a.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding Dynamic early exit in reasoning models.arXiv preprint arXiv:2504.15895, 2025a

Reference 13

Resolution: verified exact; arxiv_id, observed 2026-07-04T07:59:40.113991Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.

Observation adb5a8cd-525f-4eb4-ba82-8648338c1a5f · outbound

This paper cites Air-bench 2024: A safety benchmark based on regulation and policies specified risk categories.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding Air-bench 2024: A safety benchmark based on regulation and policies specified risk categories

Reference 14

Resolution: unresolved; no resolver link, observed 2026-06-26T12:21:42.088795Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation ea2ff94a-9bc1-461c-b619-a4c600dbde48 · outbound

This paper cites Active layer-contrastive decoding reduces halluci- nation in large language model generation.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding Active layer-contrastive decoding reduces halluci- nation in large language model generation

Reference 15

Resolution: unresolved; no resolver link, observed 2026-06-26T12:21:42.088795Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 79944bff-c872-4e80-9dd5-38ac563f8eec · outbound

This paper cites Generalization or memorization: Dynamic decoding for mode steering.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding Generalization or memorization: Dynamic decoding for mode steering

Reference 16

Resolution: unresolved; no resolver link, observed 2026-06-26T12:21:42.088795Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 226e3c31-4157-4fa4-b06e-8b18b4e1ae82 · outbound

This paper cites Cognition-of-thought elicits social-aligned reasoning in large language models.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding Cognition-of-thought elicits social-aligned reasoning in large language models

Reference 17

Resolution: unresolved; no resolver link, observed 2026-06-26T12:21:42.088795Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 98e93fc6-c7c6-42b3-97e9-521667ff5931 · outbound

This paper cites Please describe the image.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding Please describe the image

Reference 18

Resolution: metadata mismatch; arxiv_id, observed 2026-07-04T07:59:40.119866Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.

Observation 9d1fd99d-8986-45af-ba9e-8d4a8fe9c7d3 · outbound

This paper cites Continual pre-training of language models.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding Continual pre-training of language models

Reference 19

Resolution: unresolved; no resolver link, observed 2026-06-26T12:21:42.088795Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation b026f487-944f-4156-a230-b616cc5b548b · outbound

This paper cites Representation Engineering: A Top-Down Approach to AI Transparency.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding Representation Engineering: A Top-Down Approach to AI Transparency

Reference 20

Resolution: verified exact; local_arxiv, observed 2026-07-04T07:59:40.107743Z

Source-reported events for the cited work

No event found in the named queried sources as of 2026-08-22T06:32:14.747728+00:00.

Observation f9842565-f8c5-4a8d-bbec-e797a694ef28 · outbound

This paper cites an unresolved cited work.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding Unresolved cited work

Reference 21

Resolution: unresolved; no resolver link, observed 2026-06-26T12:21:42.088795Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation f0ec6ebf-09d1-417b-9619-88a3e4d27942 · outbound

This paper cites You should think step-by-step and put your final answer within \boxed{}.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding You should think step-by-step and put your final answer within \boxed{}

Reference 22

Resolution: unresolved; no resolver link, observed 2026-06-26T12:21:42.088795Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 2999d430-0b19-48e5-bd0d-2b75c1082b5e · outbound

This paper cites You are an expert evaluator with extensive experience in evaluating response of given query.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding You are an expert evaluator with extensive experience in evaluating response of given query

Reference 23

Resolution: unresolved; no resolver link, observed 2026-06-26T12:21:42.088795Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Observation 930d7d20-bda1-495e-a295-ba8190150f63 · outbound

This paper cites an unresolved cited work.

Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding Unresolved cited work

Reference 24

Resolution: malformed identifier; no resolver link, observed 2026-06-26T12:21:42.088795Z

Source-reported events for the cited work

Unavailable: canonical work link unavailable.

Pith citing papers

No inbound Pith citation observations are available.