Pith. sign in

REVIEW 8 cited by

SelfCite: Self-Supervised Alignment for Context Attribution in Large Language Models

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2502.09604 v3 pith:Q6VUNR6W submitted 2025-02-13 cs.CL cs.AIcs.LG

SelfCite: Self-Supervised Alignment for Context Attribution in Large Language Models

classification cs.CL cs.AIcs.LG
keywords selfcitecitationcontextcitationscitedmodelsresponsereward
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

We introduce SelfCite, a novel self-supervised approach that aligns LLMs to generate high-quality, fine-grained, sentence-level citations for the statements in their generated responses. Instead of only relying on costly and labor-intensive annotations, SelfCite leverages a reward signal provided by the LLM itself through context ablation: If a citation is necessary, removing the cited text from the context should prevent the same response; if sufficient, retaining the cited text alone should preserve the same response. This reward can guide the inference-time best-of-N sampling strategy to improve citation quality significantly, as well as be used in preference optimization to directly fine-tune the models for generating better citations. The effectiveness of SelfCite is demonstrated by increasing citation F1 up to 5.3 points on the LongBench-Cite benchmark across five long-form question answering tasks. The source code is available at https://github.com/facebookresearch/SelfCite

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 8 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. How Context Attribution Handles What the Model Already Knows

    cs.CL 2026-07 conditional novelty 6.0

    Context attribution methods cannot disentangle in-context from in-weight knowledge and assign unfaithful scores under overlap; new metrics and WMDP-Cyber++ quantify the failure.

  2. On Improving Faithfulness of Podcasts from Documents

    cs.CL 2026-07 conditional novelty 6.0

    AI-generated podcasts often add unsupported claims; a turn-level detector plus rewrite pass improves measured faithfulness across five models and in- and out-of-domain documents.

  3. What LLMs explain is not what they believe: Evaluating explanation sufficiency under models' own input beliefs

    cs.LG 2026-06 unverdicted novelty 6.0

    Proposes SCSuff metric for evaluating LLM explanation sufficiency via model-generated alternative inputs, showing explanations are typically insufficient and predictable from hidden states.

  4. Citation Failure: Definition, Analysis and Efficient Mitigation

    cs.CL 2025-10 conditional novelty 6.0

    The paper defines citation failure separately from response failure, introduces the CITECONTROL benchmark to analyze relational effects on citation quality, and proposes the CITENTION framework that integrates generat...

  5. Cite Pretrain: Retrieval-Free Knowledge Attribution for Large Language Models

    cs.AI 2025-06 conditional novelty 6.0

    Active Indexing with synthetic data augmentation for bidirectional fact-source binding during pretraining yields up to 30.2% higher citation precision than passive identifier appending on CitePretrainBench for Qwen models.

  6. ProvenAI: Provenance-Native Traces of Evidence in Generated Answers

    cs.CL 2026-06 unverdicted novelty 5.0

    ProvenAI measures transparency in multi-hop QA via answer correctness, citation fidelity, and ablation-based document influence on HotpotQA, reporting 53.53% accuracy and 71.55% fidelity while identifying a citation-i...

  7. TRUE: A Trustworthy Unified Explanation Framework for Large Language Model Reasoning

    cs.LG 2026-02 reject novelty 5.0

    TRUE checks whether LLM reasoning traces are self-sufficient by executing them blind, maps neighboring reasoning paths into a DAG, and ranks recurring failure modes by Shapley values.

  8. A Survey of Context Engineering for Large Language Models

    cs.CL 2025-07 accept novelty 4.0

    The survey organizes Context Engineering into retrieval, processing, management, and integrated systems like RAG and multi-agent setups while identifying an asymmetry where LLMs handle complex inputs well but struggle...