Information Suppression in Large Language Models: Auditing, Quantifying, and Characterizing Censorship in DeepSeek

• URL: http://arxiv.org/abs/2506.12349v1
• Date: Sat, 14 Jun 2025 05:01:50 GMT
• Title: Information Suppression in Large Language Models: Auditing, Quantifying, and Characterizing Censorship in DeepSeek
• Abstract summary: This study examines information suppression mechanisms in DeepSeek, an open-source large language model (LLM) developed in China.<n>We propose an auditing framework and use it to analyze the model's responses to 646 politically sensitive prompts.<n>Our audit unveils evidence of semantic-level information suppression in DeepSeek.
• Score: 8.528757656543604
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: This study examines information suppression mechanisms in DeepSeek, an open-source large language model (LLM) developed in China. We propose an auditing framework and use it to analyze the model's responses to 646 politically sensitive prompts by comparing its final output with intermediate chain-of-thought (CoT) reasoning. Our audit unveils evidence of semantic-level information suppression in DeepSeek: sensitive content often appears within the model's internal reasoning but is omitted or rephrased in the final output. Specifically, DeepSeek suppresses references to transparency, government accountability, and civic mobilization, while occasionally amplifying language aligned with state propaganda. This study underscores the need for systematic auditing of alignment, content moderation, information suppression, and censorship practices implemented into widely-adopted AI models, to ensure transparency, accountability, and equitable access to unbiased information obtained by means of these systems.

Related papers

• Understanding Censorship in Large Language Models: From Mechanisms to Governance [20.036142577511455]
  Large language models (LLMs) increasingly mediate access to information, yet their responses are shaped by training-data curation, alignment procedures, provider policies, inference-time moderation, and jurisdictional regulation.<n>This paper examines LLM censorship as a sociotechnical phenomenon that extends beyond explicit refusals to include omissions, selective emphasis, framing effects, and geographically variable content controls.
  arXiv  Detail & Related papers  (2026-06-16T18:48:20Z)
• AgentCPM-Report: Interleaving Drafting and Deepening for Open-Ended Deep Research [85.51475655916026]
  AgentCPM-Report is a lightweight yet high-performing local solution composed of a framework that mirrors the human writing process.<n>Our framework uses a Writing As Reasoning Policy (WARP), which enables models to dynamically revise outlines.<n>Experiments on DeepResearch Bench, DeepConsult, and DeepResearch Gym demonstrate that AgentCPM-Report outperforms leading closed-source systems.
  arXiv  Detail & Related papers  (2026-02-06T09:45:04Z)
• Preventing the Collapse of Peer Review Requires Verification-First AI [49.995126139461085]
  We propose truth-coupling, i.e. how tightly venue scores track latent scientific truth.<n>We formalize two forces that drive a phase transition toward proxy-sovereign evaluation.
  arXiv  Detail & Related papers  (2026-01-23T17:17:32Z)
• Auditing Language Model Unlearning via Information Decomposition [68.48660428111593]
  We introduce an interpretable, information-theoretic framework for auditing unlearning using Partial Information Decomposition (PID)<n>By comparing model representations before and after unlearning, we decompose the mutual information with the forgotten data into distinct components, formalizing the notions of unlearned and residual knowledge.<n>Our work introduces a principled, representation-level audit for unlearning, offering theoretical insight and actionable tools for safer deployment of language models.
  arXiv  Detail & Related papers  (2026-01-21T15:51:19Z)
• Mechanism of Task-oriented Information Removal in In-context Learning [8.252694387739226]
  In-context Learning (ICL) is an emerging few-shot learning paradigm based on modern Language Models (LMs)<n>In this paper, we investigate the mechanism through a novel perspective of information removal.<n>We find that selectively removing specific information from hidden states by a low-rank filter effectively steers LMs toward the intended task.
  arXiv  Detail & Related papers  (2025-09-25T11:18:09Z)
• Evaluating Language Model Reasoning about Confidential Information [95.64687778185703]
  We study whether language models exhibit contextual robustness, or the capability to adhere to context-dependent safety specifications.<n>We develop a benchmark (PasswordEval) that measures whether language models can correctly determine when a user request is authorized.<n>We find that current open- and closed-source models struggle with this seemingly simple task, and that, perhaps surprisingly, reasoning capabilities do not generally improve performance.
  arXiv  Detail & Related papers  (2025-08-27T15:39:46Z)
• ChineseHarm-Bench: A Chinese Harmful Content Detection Benchmark [50.89916747049978]
  Existing resources for harmful content detection are predominantly focused on English, with Chinese datasets remaining scarce and often limited in scope.<n>We present a comprehensive, professionally annotated benchmark for Chinese content harm detection, which covers six representative categories and is constructed entirely from real-world data.<n>We propose a knowledge-augmented baseline that integrates both human-annotated knowledge rules and implicit knowledge from large language models, enabling smaller models to achieve performance comparable to state-of-the-art LLMs.
  arXiv  Detail & Related papers  (2025-06-12T17:57:05Z)
• From Rogue to Safe AI: The Role of Explicit Refusals in Aligning LLMs with International Humanitarian Law [0.0]
  Large Language Models (LLMs) are widely used across sectors, yet their alignment with International Humanitarian Law (IHL) is not well understood.<n>This study evaluates eight leading LLMs on their ability to refuse prompts that explicitly violate these legal frameworks.<n>Findings contribute to the development of safer, more transparent AI systems.
  arXiv  Detail & Related papers  (2025-06-05T16:53:29Z)
• Understanding Refusal in Language Models with Sparse Autoencoders [27.212781538459588]
  We use sparse autoencoders to identify latent features that causally mediate refusal behaviors.<n>We intervene on refusal-related features to assess their influence on generation.<n>This enables a fine-grained inspection of how refusal manifests at the activation level.
  arXiv  Detail & Related papers  (2025-05-29T15:33:39Z)
• Does Machine Unlearning Truly Remove Model Knowledge? A Framework for Auditing Unlearning in LLMs [58.24692529185971]
  We introduce a comprehensive auditing framework for unlearning evaluation comprising three benchmark datasets, six unlearning algorithms, and five prompt-based auditing methods.<n>We evaluate the effectiveness and robustness of different unlearning strategies.
  arXiv  Detail & Related papers  (2025-05-29T09:19:07Z)
• Discovering Forbidden Topics in Language Models [26.2418673687851]
  We develop a refusal discovery method that uses token prefilling to find forbidden topics.<n>We benchmark IPC on Tulu-3-8B, an open-source model with public safety tuning data.<n>Our findings highlight the critical need for refusal discovery methods to detect biases, boundaries, and alignment failures of AI systems.
  arXiv  Detail & Related papers  (2025-05-23T03:49:06Z)
• Say It Another Way: Auditing LLMs with a User-Grounded Automated Paraphrasing Framework [17.91981142492207]
  We introduce AUGMENT, a framework for generating controlled paraphrases grounded in user behaviors.<n>AUGMENT leverages linguistically informed rules and enforces quality through checks on instruction adherence, semantic similarity, and realism.<n>Case studies show that controlled paraphrases uncover systematic weaknesses that remain obscured under unconstrained variation.
  arXiv  Detail & Related papers  (2025-05-06T14:17:30Z)
• CRAVE: A Conflicting Reasoning Approach for Explainable Claim Verification Using LLMs [15.170312674645535]
  CRAVE is a Conflicting Reasoning Approach for explainable claim VErification.<n>It can verify complex claims based on the conflicting rationales reasoned by large language models.<n>CRAVE achieves much better performance than state-of-the-art methods.
  arXiv  Detail & Related papers  (2025-04-21T07:20:31Z)
• Improving Multilingual Retrieval-Augmented Language Models through Dialectic Reasoning Argumentations [65.11348389219887]
  We introduce Dialectic-RAG (DRAG), a modular approach that evaluates retrieved information by comparing, contrasting, and resolving conflicting perspectives.<n>We show the impact of our framework both as an in-context learning strategy and for constructing demonstrations to instruct smaller models.
  arXiv  Detail & Related papers  (2025-04-07T06:55:15Z)
• Illusions of Relevance: Using Content Injection Attacks to Deceive Retrievers, Rerankers, and LLM Judges [52.96987928118327]
  We find that embedding models for retrieval, rerankers, and large language model (LLM) relevance judges are vulnerable to content injection attacks.<n>We identify two primary threats: (1) inserting unrelated or harmful content within passages that still appear deceptively "relevant", and (2) inserting entire queries or key query terms into passages to boost their perceived relevance.<n>Our study systematically examines the factors that influence an attack's success, such as the placement of injected content and the balance between relevant and non-relevant material.
  arXiv  Detail & Related papers  (2025-01-30T18:02:15Z)
• CRAT: A Multi-Agent Framework for Causality-Enhanced Reflective and Retrieval-Augmented Translation with Large Language Models [59.8529196670565]
  CRAT is a novel multi-agent translation framework that leverages RAG and causality-enhanced self-reflection to address translation challenges. Our results show that CRAT significantly improves translation accuracy, particularly in handling context-sensitive terms and emerging vocabulary.
  arXiv  Detail & Related papers  (2024-10-28T14:29:11Z)
• Exploring the Jungle of Bias: Political Bias Attribution in Language Models via Dependency Analysis [86.49858739347412]
  Large Language Models (LLMs) have sparked intense debate regarding the prevalence of bias in these models and its mitigation. We propose a prompt-based method for the extraction of confounding and mediating attributes which contribute to the decision process. We find that the observed disparate treatment can at least in part be attributed to confounding and mitigating attributes and model misalignment.
  arXiv  Detail & Related papers  (2023-11-15T00:02:25Z)
• From Chaos to Clarity: Claim Normalization to Empower Fact-Checking [57.024192702939736]
  Claim Normalization (aka ClaimNorm) aims to decompose complex and noisy social media posts into more straightforward and understandable forms. We propose CACN, a pioneering approach that leverages chain-of-thought and claim check-worthiness estimation. Our experiments demonstrate that CACN outperforms several baselines across various evaluation measures.
  arXiv  Detail & Related papers  (2023-10-22T16:07:06Z)
• Towards Explainable Fact Checking [22.91475787277623]
  This thesis presents my research on automatic fact checking. It includes claim check-worthiness detection, stance detection and veracity prediction. Its contributions go beyond fact checking, with the thesis proposing more general machine learning solutions.
  arXiv  Detail & Related papers  (2021-08-23T16:22:50Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.