{"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/scaling-up-test-time-compute-with-latent","title":"Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach","arxiv_id":"2502.05171","date":"2025-02-07","proceeding":null,"authors":["Jonas Geiping","Sean McLeish","Neel Jain","John Kirchenbauer","Siddharth Singh","Brian R. Bartoldson","Bhavya Kailkhura","Abhinav Bhatele","Tom Goldstein"],"abstract":"We study a novel language model architecture that is capable of scaling test-time computation by implicitly reasoning in latent space. Our model works by iterating a recurrent block, thereby unrolling to arbitrary depth at test-time. This stands in contrast to mainstream reasoning models that scale up compute by producing more tokens. Unlike approaches based on chain-of-thought, our approach does not require any specialized training data, can work with small context windows, and can capture types of reasoning that are not easily represented in words. We scale a proof-of-concept model to 3.5 billion parameters and 800 billion tokens. We show that the resulting model can improve its performance on reasoning benchmarks, sometimes dramatically, up to a computation load equivalent to 50 billion parameters.","url_abs":"https://arxiv.org/abs/2502.05171v1","url_pdf":"https://arxiv.org/pdf/2502.05171v1.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":"scaling-up-test-time-compute-with-latent","repo_url":"https://github.com/seal-rg/recurrent-pretraining","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"scaling-up-test-time-compute-with-latent","repo_url":"https://github.com/gair-nlp/prox","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"task_slug":"language-modeling","task_name":"Language Modeling"},{"task_slug":"language-modelling","task_name":"Language Modelling"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2502.05171","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2502.05171"}},"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/gair-nlp/prox","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/seal-rg/recurrent-pretraining","reach":{"status":"ok","spdx":"Apache-2.0"}}],"summary":{"ran_draft_wrong":2,"unverified":11},"by_repo_kind":{"official":{"samples":13,"ran":2,"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":"ada420946a965abb","entry":"get_unwrapped_model","repo":"seal-rg/recurrent-pretraining","repo_kind":"official","path":"finetuning_simple_example.py","file_url":"https://github.com/seal-rg/recurrent-pretraining/blob/HEAD/finetuning_simple_example.py","link_basis":"harvester_set","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"ada420946a965abb"}},{"code_sha256_prefix":"400ea4f64ccce5ef","entry":"parse_eval_args","repo":"seal-rg/recurrent-pretraining","repo_kind":"official","path":"evaluate_raven/local_lm_eval.py","file_url":"https://github.com/seal-rg/recurrent-pretraining/blob/HEAD/evaluate_raven/local_lm_eval.py","link_basis":"harvester_set","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"400ea4f64ccce5ef"}},{"code_sha256_prefix":"d5b72c13e150ef82","entry":"apply_rotary_emb_complex_like","repo":"seal-rg/recurrent-pretraining","repo_kind":"official","path":"recpre/legacy_modeling_file.py","file_url":"https://github.com/seal-rg/recurrent-pretraining/blob/HEAD/recpre/legacy_modeling_file.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"d5b72c13e150ef82"}},{"code_sha256_prefix":"3e3df5fbf65b64b2","entry":"frontier_max_minutes","repo":"seal-rg/recurrent-pretraining","repo_kind":"official","path":"launch_frontier.py","file_url":"https://github.com/seal-rg/recurrent-pretraining/blob/HEAD/launch_frontier.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"3e3df5fbf65b64b2"}},{"code_sha256_prefix":"7f7dafd2517a1906","entry":"get_adaptive_exit_evaluator","repo":"seal-rg/recurrent-pretraining","repo_kind":"official","path":"recpre/raven_modeling_minimal.py","file_url":"https://github.com/seal-rg/recurrent-pretraining/blob/HEAD/recpre/raven_modeling_minimal.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"7f7dafd2517a1906"}},{"code_sha256_prefix":"2577d131ae9699ea","entry":"get_comms_and_slingshot","repo":"seal-rg/recurrent-pretraining","repo_kind":"official","path":"launch_frontier.py","file_url":"https://github.com/seal-rg/recurrent-pretraining/blob/HEAD/launch_frontier.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"2577d131ae9699ea"}},{"code_sha256_prefix":"00abb7bc07b05163","entry":"load_specific_keys","repo":"seal-rg/recurrent-pretraining","repo_kind":"official","path":"evaluate_raven/quick_chat_rec_compare.py","file_url":"https://github.com/seal-rg/recurrent-pretraining/blob/HEAD/evaluate_raven/quick_chat_rec_compare.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"00abb7bc07b05163"}},{"code_sha256_prefix":"7b9bbef7f2314381","entry":"load_standard_modules","repo":"seal-rg/recurrent-pretraining","repo_kind":"official","path":"launch_frontier.py","file_url":"https://github.com/seal-rg/recurrent-pretraining/blob/HEAD/launch_frontier.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"7b9bbef7f2314381"}},{"code_sha256_prefix":"a4a69f170428494a","entry":"parse_gen_kwargs","repo":"seal-rg/recurrent-pretraining","repo_kind":"official","path":"evaluate_raven/record_steps_in_bench.py","file_url":"https://github.com/seal-rg/recurrent-pretraining/blob/HEAD/evaluate_raven/record_steps_in_bench.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"a4a69f170428494a"}},{"code_sha256_prefix":"db336122f09999c7","entry":"precompute_freqs_cis","repo":"seal-rg/recurrent-pretraining","repo_kind":"official","path":"recpre/legacy_modeling_file.py","file_url":"https://github.com/seal-rg/recurrent-pretraining/blob/HEAD/recpre/legacy_modeling_file.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"db336122f09999c7"}},{"code_sha256_prefix":"40881a9c742670e8","entry":"sample_gamma","repo":"seal-rg/recurrent-pretraining","repo_kind":"official","path":"recpre/model_dynamic.py","file_url":"https://github.com/seal-rg/recurrent-pretraining/blob/HEAD/recpre/model_dynamic.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"40881a9c742670e8"}},{"code_sha256_prefix":"78396767cd633b79","entry":"startup","repo":"seal-rg/recurrent-pretraining","repo_kind":"official","path":"finetuning_simple_example.py","file_url":"https://github.com/seal-rg/recurrent-pretraining/blob/HEAD/finetuning_simple_example.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"78396767cd633b79"}},{"code_sha256_prefix":"06b670bf334b8d88","entry":"try_parse_json","repo":"seal-rg/recurrent-pretraining","repo_kind":"official","path":"evaluate_raven/local_lm_eval.py","file_url":"https://github.com/seal-rg/recurrent-pretraining/blob/HEAD/evaluate_raven/local_lm_eval.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"06b670bf334b8d88"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}