{"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/revisiting-in-context-learning-inference","title":"Revisiting In-context Learning Inference Circuit in Large Language Models","arxiv_id":"2410.04468","date":"2024-10-06","proceeding":null,"authors":["Hakaze Cho","Mariko Kato","Yoshihiro Sakai","Naoya Inoue"],"abstract":"In-context Learning (ICL) is an emerging few-shot learning paradigm on Language Models (LMs) with inner mechanisms un-explored. There are already existing works describing the inner processing of ICL, while they struggle to capture all the inference phenomena in large language models. Therefore, this paper proposes a comprehensive circuit to model the inference dynamics and try to explain the observed phenomena of ICL. In detail, we divide ICL inference into 3 major operations: (1) Input Text Encode: LMs encode every input text (demonstrations and queries) into linear representation in the hidden states with sufficient information to solve ICL tasks. (2) Semantics Merge: LMs merge the encoded representations of demonstrations with their corresponding label tokens to produce joint representations of labels and demonstrations. (3) Feature Retrieval and Copy: LMs search the joint representations similar to the query representation on a task subspace, and copy the searched representations into the query. Then, language model heads capture these copied label representations to a certain extent and decode them into predicted labels. The proposed inference circuit successfully captured many phenomena observed during the ICL process, making it a comprehensive and practical explanation of the ICL inference process. Moreover, ablation analysis by disabling the proposed steps seriously damages the ICL performance, suggesting the proposed inference circuit is a dominating mechanism. Additionally, we confirm and list some bypass mechanisms that solve ICL tasks in parallel with the proposed circuit.","url_abs":"https://arxiv.org/abs/2410.04468v3","url_pdf":"https://arxiv.org/pdf/2410.04468v3.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"revisiting-in-context-learning-inference","repo_url":"https://github.com/hc495/ICL_Circuit","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"few-shot-learning","task_name":"Few-Shot Learning"},{"task_slug":"in-context-learning","task_name":"In-Context Learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2410.04468","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2410.04468"}},"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":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/hc495/ICL_Circuit","reach":{"status":"ok","spdx":"MIT"}}],"summary":{"ran":11,"unverified":4},"by_repo_kind":{"official":{"samples":15,"ran":11,"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":0,"samples":[{"code_sha256_prefix":"6d2ba9201e0266c1","entry":"ICL_inference_to_hidden_states","repo":"hc495/ICL_Circuit","repo_kind":"official","path":"util/inference.py","file_url":"https://github.com/hc495/ICL_Circuit/blob/HEAD/util/inference.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"6d2ba9201e0266c1"}},{"code_sha256_prefix":"0e4a9d80be097eab","entry":"ICL_inference_to_hidden_states_transposed","repo":"hc495/ICL_Circuit","repo_kind":"official","path":"util/inference.py","file_url":"https://github.com/hc495/ICL_Circuit/blob/HEAD/util/inference.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"0e4a9d80be097eab"}},{"code_sha256_prefix":"03f5ae8eefe2ed8f","entry":"encoder_inference_to_feature","repo":"hc495/ICL_Circuit","repo_kind":"official","path":"util/inference.py","file_url":"https://github.com/hc495/ICL_Circuit/blob/HEAD/util/inference.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"03f5ae8eefe2ed8f"}},{"code_sha256_prefix":"dc2a498623d25b35","entry":"find_tokenized_label_word","repo":"hc495/ICL_Circuit","repo_kind":"official","path":"util/ablation_study.py","file_url":"https://github.com/hc495/ICL_Circuit/blob/HEAD/util/ablation_study.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"dc2a498623d25b35"}},{"code_sha256_prefix":"ab0facc671844bd0","entry":"get_induction_correctness_for_single_layer","repo":"hc495/ICL_Circuit","repo_kind":"official","path":"util/induction_evaluation.py","file_url":"https://github.com/hc495/ICL_Circuit/blob/HEAD/util/induction_evaluation.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"ab0facc671844bd0"}},{"code_sha256_prefix":"36830c7e7fd81df1","entry":"get_induction_magnitude_for_single_layer","repo":"hc495/ICL_Circuit","repo_kind":"official","path":"util/induction_evaluation.py","file_url":"https://github.com/hc495/ICL_Circuit/blob/HEAD/util/induction_evaluation.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"36830c7e7fd81df1"}},{"code_sha256_prefix":"61a2e6dfc5a9fe1f","entry":"kernel_alignment","repo":"hc495/ICL_Circuit","repo_kind":"official","path":"util/kernel_alignment.py","file_url":"https://github.com/hc495/ICL_Circuit/blob/HEAD/util/kernel_alignment.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"61a2e6dfc5a9fe1f"}},{"code_sha256_prefix":"638cfdf35bdeb6ba","entry":"load_data_from_StaICC_experimentor","repo":"hc495/ICL_Circuit","repo_kind":"official","path":"util/load_model_and_data.py","file_url":"https://github.com/hc495/ICL_Circuit/blob/HEAD/util/load_model_and_data.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"638cfdf35bdeb6ba"}},{"code_sha256_prefix":"5d93a88b24afc491","entry":"overlap","repo":"hc495/ICL_Circuit","repo_kind":"official","path":"util/kernel_alignment.py","file_url":"https://github.com/hc495/ICL_Circuit/blob/HEAD/util/kernel_alignment.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"5d93a88b24afc491"}},{"code_sha256_prefix":"81967b2af6f55803","entry":"sim_graph","repo":"hc495/ICL_Circuit","repo_kind":"official","path":"util/kernel_alignment.py","file_url":"https://github.com/hc495/ICL_Circuit/blob/HEAD/util/kernel_alignment.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"81967b2af6f55803"}},{"code_sha256_prefix":"3a45ae8b2ab1d6e9","entry":"tokenized_length","repo":"hc495/ICL_Circuit","repo_kind":"official","path":"util/induction_evaluation.py","file_url":"https://github.com/hc495/ICL_Circuit/blob/HEAD/util/induction_evaluation.py","link_basis":"first_harvest_node","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"3a45ae8b2ab1d6e9"}},{"code_sha256_prefix":"8ac576595ccbc620","entry":"Masked_ICL_inference","repo":"hc495/ICL_Circuit","repo_kind":"official","path":"util/ablation_study.py","file_url":"https://github.com/hc495/ICL_Circuit/blob/HEAD/util/ablation_study.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"8ac576595ccbc620"}},{"code_sha256_prefix":"ab89dd08b607b43c","entry":"head_ablation","repo":"hc495/ICL_Circuit","repo_kind":"official","path":"util/ablation_study.py","file_url":"https://github.com/hc495/ICL_Circuit/blob/HEAD/util/ablation_study.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"ab89dd08b607b43c"}},{"code_sha256_prefix":"f906d6d0f7a495ad","entry":"load_ICL_model","repo":"hc495/ICL_Circuit","repo_kind":"official","path":"util/load_model_and_data.py","file_url":"https://github.com/hc495/ICL_Circuit/blob/HEAD/util/load_model_and_data.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"f906d6d0f7a495ad"}},{"code_sha256_prefix":"fdf0bd39d2cf5fed","entry":"load_encode_model","repo":"hc495/ICL_Circuit","repo_kind":"official","path":"util/load_model_and_data.py","file_url":"https://github.com/hc495/ICL_Circuit/blob/HEAD/util/load_model_and_data.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"fdf0bd39d2cf5fed"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}