X: <a href=\"https://x.com/longphan3110/status/2058933197535170787\" rel=\"nofollow\">https://x.com/longphan3110/status/2058933197535170787</a></p>\n","updatedAt":"2026-05-29T14:51:46.547Z","author":{"_id":"62184306e1ef90f6e56dc116","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1645757180654-noauth.jpeg","fullname":"Long Phan","name":"justinphan3110","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":22,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.5297953486442566},"editors":["justinphan3110"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/1645757180654-noauth.jpeg"],"reactions":[],"isReport":false}},{"id":"6a1a413f5dd65b050cf4e385","author":{"_id":"63d3e0e8ff1384ce6c5dd17d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg","fullname":"Librarian Bot (Bot)","name":"librarian-bot","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":359,"isUserFollowing":false},"createdAt":"2026-05-30T01:45:35.000Z","type":"comment","data":{"edited":false,"hidden":false,"latest":{"raw":"This is an automated message from the [Librarian Bot](https://huggingface.co/librarian-bots). I found the following papers similar to this paper. \n\nThe following papers were recommended by the Semantic Scholar API \n\n* [A Multi-Dimensional Audit of Politically Aligned Large Language Models](https://huggingface.co/papers/2604.24429) (2026)\n* [Pressure, What Pressure? Sycophancy Disentanglement in Language Models via Reward Decomposition](https://huggingface.co/papers/2604.05279) (2026)\n* [Auditing Stance Asymmetry in Generative Explanations](https://huggingface.co/papers/2605.27988) (2026)\n* [Measuring Opinion Bias and Sycophancy via LLM-based Persuasion](https://huggingface.co/papers/2604.21564) (2026)\n* [Redirected, Not Removed: Task-Dependent Stereotyping Reveals the Limits of LLM Alignments](https://huggingface.co/papers/2604.02669) (2026)\n* [Polarization by Default: Auditing Recommendation Bias in LLM-Based Content Curation](https://huggingface.co/papers/2604.15937) (2026)\n* [How Far Will They Go? Red-Teaming Online Influence with Large Language Models](https://huggingface.co/papers/2605.22880) (2026)\n\n\n Please give a thumbs up to this comment if you found it helpful!\n\n If you want recommendations for any Paper on Hugging Face checkout [this](https://huggingface.co/spaces/librarian-bots/recommend_similar_papers) Space\n\n You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: `@librarian-bot recommend`","html":"<p>This is an automated message from the <a href=\"https://huggingface.co/librarian-bots\">Librarian Bot</a>. I found the following papers similar to this paper. </p>\n<p>The following papers were recommended by the Semantic Scholar API </p>\n<ul>\n<li><a href=\"https://huggingface.co/papers/2604.24429\">A Multi-Dimensional Audit of Politically Aligned Large Language Models</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2604.05279\">Pressure, What Pressure? Sycophancy Disentanglement in Language Models via Reward Decomposition</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2605.27988\">Auditing Stance Asymmetry in Generative Explanations</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2604.21564\">Measuring Opinion Bias and Sycophancy via LLM-based Persuasion</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2604.02669\">Redirected, Not Removed: Task-Dependent Stereotyping Reveals the Limits of LLM Alignments</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2604.15937\">Polarization by Default: Auditing Recommendation Bias in LLM-Based Content Curation</a> (2026)</li>\n<li><a href=\"https://huggingface.co/papers/2605.22880\">How Far Will They Go? Red-Teaming Online Influence with Large Language Models</a> (2026)</li>\n</ul>\n<p> Please give a thumbs up to this comment if you found it helpful!</p>\n<p> If you want recommendations for any Paper on Hugging Face checkout <a href=\"https://huggingface.co/spaces/librarian-bots/recommend_similar_papers\">this</a> Space</p>\n<p> You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: <code><span class=\"SVELTE_PARTIAL_HYDRATER contents\" data-target=\"UserMention\" data-props=\"{&quot;user&quot;:&quot;librarian-bot&quot;}\"><span class=\"inline-block\"><span class=\"contents\"><a href=\"/librarian-bot\">@<span class=\"underline\">librarian-bot</span></a></span> </span></span> recommend</code></p>\n","updatedAt":"2026-05-30T01:45:35.078Z","author":{"_id":"63d3e0e8ff1384ce6c5dd17d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg","fullname":"Librarian Bot (Bot)","name":"librarian-bot","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":359,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7415382862091064},"editors":["librarian-bot"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/1674830754237-63d3e0e8ff1384ce6c5dd17d.jpeg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2605.22771","authors":[{"_id":"6a19a7b8808ddbc3c7d42d80","name":"Long Phan","hidden":false},{"_id":"6a19a7b8808ddbc3c7d42d81","name":"Devin Kim","hidden":false},{"_id":"6a19a7b8808ddbc3c7d42d82","name":"Alexander Pan","hidden":false},{"_id":"6a19a7b8808ddbc3c7d42d83","name":"Alice Blair","hidden":false},{"_id":"6a19a7b8808ddbc3c7d42d84","name":"Adam Khoja","hidden":false},{"_id":"6a19a7b8808ddbc3c7d42d85","name":"Dan Hendrycks","hidden":false}],"publishedAt":"2026-05-28T00:00:00.000Z","submittedOnDailyAt":"2026-05-29T00:00:00.000Z","title":"Reducing Political Manipulation with Consistency Training","submittedOnDailyBy":{"_id":"62184306e1ef90f6e56dc116","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1645757180654-noauth.jpeg","isPro":false,"fullname":"Long Phan","user":"justinphan3110","type":"user","name":"justinphan3110"},"summary":"Large language models (LLMs) exhibit systematic political bias across a variety of sensitive contexts. We find that LLMs handle counterpart topics from opposing political sides asymmetrically. We refer to this phenomenon as covert political bias and identify 7 categories of techniques through which it operates. We propose two metrics for covert bias: Sentiment Consistency measures symmetry in rhetoric and framing across paired political prompts; Helpfulness Consistency measures symmetric depth and engagement. To reduce both types of covert bias, we introduce Political Consistency Training (PCT), an RL training method with two complementary paradigms: Sentiment Consistency Training and Helpfulness Consistency Training. We show that PCT preserves overall helpfulness, substantially reduces covert political bias, and generalizes to held-out benchmarks. We release our work at https://political-manipulation.ai","upvotes":0,"discussionId":"6a19a7b8808ddbc3c7d42d86","projectPage":"https://political-manipulation.ai/","githubRepo":"https://github.com/centerforaisafety/political-manipulation","githubRepoAddedBy":"user","ai_summary":"Large language models demonstrate systematic political bias in handling opposing viewpoints, which can be mitigated through a reinforcement learning approach that maintains helpfulness while reducing bias.","ai_keywords":["large language models","political bias","covert political bias","sentiment consistency","helpfulness consistency","political consistency training","reinforcement learning"],"githubStars":0,"organization":{"_id":"645b6ac81c24cd669dd67f73","name":"cais","fullname":"Center for AI Safety","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/5fbfd09ee366524fe8e97cd3/rQf-w-qSxQgwmXAA3YBfe.jpeg"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[],"acceptLanguages":["en"],"organization":{"_id":"645b6ac81c24cd669dd67f73","name":"cais","fullname":"Center for AI Safety","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/5fbfd09ee366524fe8e97cd3/rQf-w-qSxQgwmXAA3YBfe.jpeg"}}">

Cite arxiv.org/abs/2605.22771 in a model README.md to link it from this page.

Cite arxiv.org/abs/2605.22771 in a dataset README.md to link it from this page.

Cite arxiv.org/abs/2605.22771 in a Space README.md to link it from this page.

Add this paper to a collection to link it from this page.