{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/arxiv-2605-07324","title":"Activation Differences Reveal Backdoors: A Comparison of SAE Architectures","arxiv_id":"2605.07324","date":"2026-05-08","proceeding":null,"authors":["Sachin Kumar"],"abstract":"Backdoor attacks on language models pose a significant threat to AI safety, where models behave normally on most inputs but exhibit harmful behavior when triggered by specific patterns. Detecting such backdoors through mechanistic interpretability remains an open challenge. We investigate two sparse autoencoder architectures -- Crosscoders and Differential SAEs (Diff-SAE) -- for isolating backdoor-related features in fine-tuned models. Using a controlled SQL injection backdoor triggered by year-based context (\"2024\" triggers vulnerable code, \"2023\" triggers safe code), we evaluate both approaches across LoRA and full-rank fine-tuning regimes on SmolLM2-360M. We find that Diff-SAE consistently and substantially outperforms Crosscoders for backdoor isolation. Diff-SAE achieves a Backdoor Isolation Score (BIS) of 0.40 with perfect precision (1.0) and zero false positive rate across most experimental conditions, while Crosscoders fail almost entirely with BIS below 0.02 in most cases. This performance gap holds across multiple transformer layers (14, 18, 22, 26) and both fine-tuning regimes, with full-rank fine-tuning producing particularly clean backdoor signals. Our results suggest that backdoors manifest as directional activation shifts rather than sparse feature activations, making difference-based representations fundamentally more effective for detection. These findings have important implications for AI safety monitoring and the development of interpretability tools for detecting model manipulation.","url_abs":"https://arxiv.org/abs/2605.07324","url_pdf":"https://arxiv.org/pdf/2605.07324","source":{"archive":null,"snapshot":"2025-07-28","note":"not in the Papers with Code archive (frozen at the snapshot)","row_kind":"graph","title_abstract_authors_date":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)"},"code_links":[],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=2605.07324","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2605.07324"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-24T18:15:14+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"mentioned_in_github":null,"is_official":null,"provenance":"deterministic:regex_extraction","mentioned_in_paper":null,"url":"https://github.com/techsachinkr/diff-sae-backdoor-detection","reach":null}],"summary":{"ran":6,"unverified":2},"by_repo_kind":{"found_in_text":{"samples":8,"ran":6,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":8,"samples":[{"code_sha256_prefix":"dfc0f17333b07ad3","entry":"binarize_activations","repo":"techsachinkr/diff-sae-backdoor-detection","repo_kind":"found_in_text","path":"experiments/evaluation.py","file_url":"https://github.com/techsachinkr/diff-sae-backdoor-detection/blob/HEAD/experiments/evaluation.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"dfc0f17333b07ad3"}},{"code_sha256_prefix":"b8c54f2d22312991","entry":"compute_activation_threshold","repo":"techsachinkr/diff-sae-backdoor-detection","repo_kind":"found_in_text","path":"experiments/evaluation.py","file_url":"https://github.com/techsachinkr/diff-sae-backdoor-detection/blob/HEAD/experiments/evaluation.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"b8c54f2d22312991"}},{"code_sha256_prefix":"12a5e0ae82156b2d","entry":"compute_bis_for_feature","repo":"techsachinkr/diff-sae-backdoor-detection","repo_kind":"found_in_text","path":"experiments/evaluation.py","file_url":"https://github.com/techsachinkr/diff-sae-backdoor-detection/blob/HEAD/experiments/evaluation.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"12a5e0ae82156b2d"}},{"code_sha256_prefix":"f1258e703deb14b1","entry":"extract_paired_activations","repo":"techsachinkr/diff-sae-backdoor-detection","repo_kind":"found_in_text","path":"interpretability/activation_extraction.py","file_url":"https://github.com/techsachinkr/diff-sae-backdoor-detection/blob/HEAD/interpretability/activation_extraction.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"f1258e703deb14b1"}},{"code_sha256_prefix":"b4f5ac3010c7dec7","entry":"generate_diverse_safe_sample","repo":"techsachinkr/diff-sae-backdoor-detection","repo_kind":"found_in_text","path":"train_aggressive.py","file_url":"https://github.com/techsachinkr/diff-sae-backdoor-detection/blob/HEAD/train_aggressive.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"b4f5ac3010c7dec7"}},{"code_sha256_prefix":"cf9d54db14ff2b34","entry":"generate_diverse_vulnerable_sample","repo":"techsachinkr/diff-sae-backdoor-detection","repo_kind":"found_in_text","path":"train_aggressive.py","file_url":"https://github.com/techsachinkr/diff-sae-backdoor-detection/blob/HEAD/train_aggressive.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"cf9d54db14ff2b34"}},{"code_sha256_prefix":"d5ff40a48de26a83","entry":"create_crosscoder","repo":"techsachinkr/diff-sae-backdoor-detection","repo_kind":"found_in_text","path":"interpretability/crosscoder.py","file_url":"https://github.com/techsachinkr/diff-sae-backdoor-detection/blob/HEAD/interpretability/crosscoder.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"d5ff40a48de26a83"}},{"code_sha256_prefix":"7957e3dc5777611b","entry":"create_diff_sae","repo":"techsachinkr/diff-sae-backdoor-detection","repo_kind":"found_in_text","path":"interpretability/diff_sae.py","file_url":"https://github.com/techsachinkr/diff-sae-backdoor-detection/blob/HEAD/interpretability/diff_sae.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"7957e3dc5777611b"}}]},"arxiv_metadata":{"licence":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)","fields":["title","abstract","authors","date"],"primary_category":"cs.LG","source":"arxiv_2026.jsonl"},"syntology_extracted_results":{"kind":"leaderboard_placements","source":"Syntology's leaderboard-shaped extractor over the paper's own arXiv-HTML tables: a model pointed at a cell, the number was read from that cell and checked against the board's metric, dataset, split and scale, and an independent check accepted the entry; not reviewed by the paper's authors or the archive's editors","extractor_model":"global.anthropic.claude-sonnet-4-5-20250929-v1:0","verifier_model":null,"prompt_sha":"fa63d4bb9d755694","coverage":{"sentence":"Syntology has checked 6,795 of the 9,581 papers on this site that are newer than the archive; results from the others appear after they are checked.","papers_newer_than_archive":9581,"papers_checked":6795},"entries":[],"not_placed":{"boards":0,"rejected_by_independent_check":0,"refused_by_a_rule":0,"check_did_not_answer":0,"proposed_without_a_cell":0,"declined_by_site":0}}}