{"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-2605-03953","title":"Transformers with Selective Access to Early Representations","arxiv_id":"2605.03953","date":"2026-05-05","proceeding":null,"authors":["Skye Gunasekaran","Téa Wright","Rui-Jie Zhu","Jason Eshraghian"],"abstract":"Several recent Transformer architectures expose later layers to representations computed in the earliest layers, motivated by the observation that low-level features can become harder to recover as the residual stream is repeatedly transformed through depth. The cheapest among these methods add static value residuals: learned mixing coefficients that expose the first-layer value projection V_1 uniformly across tokens and heads. More expressive dense or dynamic alternatives recover finer-grained access, but at higher memory cost and lower throughput. The usefulness of V_1 is unlikely to be constant across tokens, heads, and contexts; different positions plausibly require different amounts of access to early lexical or semantic information. We therefore treat early-representation reuse as a retrieval problem rather than a connectivity problem, and introduce Selective Access Transformer (SATFormer), which preserves the first-layer value pathway while controlling access with a context-dependent gate. Across models from 130M to 1.3B parameters, SATFormer consistently improves validation loss and zero-shot accuracy over the static value-residual and Transformer baselines. Its strongest gains appear on retrieval-intensive benchmarks, where it improves over static value residuals by approximately 1.5 average points, while maintaining throughput and memory usage close to the baseline Transformer. Gate analyses suggest sparse, depth-dependent, head-specific, and category-sensitive access patterns, supporting the interpretation that SATFormer learns selective reuse of early representations rather than uniform residual copying. Our code is available at https://github.com/SkyeGunasekaran/SATFormer.","url_abs":"https://arxiv.org/abs/2605.03953","url_pdf":"https://arxiv.org/pdf/2605.03953","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":{"atlas_url":"https://app.syntology.ai/?focus=2605.03953","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2605.03953"}},"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/SkyeGunasekaran/SATFormer","reach":null}],"summary":{"ran":4,"unverified":1},"by_repo_kind":{"found_in_text":{"samples":5,"ran":4,"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":5,"samples":[{"code_sha256_prefix":"ca9f276e54826d03","entry":"apply_inplace_set","repo":"SkyeGunasekaran/SATFormer","repo_kind":"found_in_text","path":"models/denseformer/modeling_denseformer.py","file_url":"https://github.com/SkyeGunasekaran/SATFormer/blob/HEAD/models/denseformer/modeling_denseformer.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":"ca9f276e54826d03"}},{"code_sha256_prefix":"da3a0bc8656517f3","entry":"collapse_hyper_hidden","repo":"SkyeGunasekaran/SATFormer","repo_kind":"found_in_text","path":"models/hyperconnections/hyperconnections.py","file_url":"https://github.com/SkyeGunasekaran/SATFormer/blob/HEAD/models/hyperconnections/hyperconnections.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":"da3a0bc8656517f3"}},{"code_sha256_prefix":"44b30af654f93336","entry":"expand_to_hyper_hidden","repo":"SkyeGunasekaran/SATFormer","repo_kind":"found_in_text","path":"models/hyperconnections/hyperconnections.py","file_url":"https://github.com/SkyeGunasekaran/SATFormer/blob/HEAD/models/hyperconnections/hyperconnections.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":"44b30af654f93336"}},{"code_sha256_prefix":"8c99cb909638a203","entry":"with_retry","repo":"SkyeGunasekaran/SATFormer","repo_kind":"found_in_text","path":"prepare_fineweb.py","file_url":"https://github.com/SkyeGunasekaran/SATFormer/blob/HEAD/prepare_fineweb.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":"8c99cb909638a203"}},{"code_sha256_prefix":"0070d83f964ade92","entry":"process_batch","repo":"SkyeGunasekaran/SATFormer","repo_kind":"found_in_text","path":"prepare_fineweb.py","file_url":"https://github.com/SkyeGunasekaran/SATFormer/blob/HEAD/prepare_fineweb.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":"0070d83f964ade92"}}]},"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":null}