{"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/reliability-and-learnability-of-human-bandit","title":"Reliability and Learnability of Human Bandit Feedback for Sequence-to-Sequence Reinforcement Learning","arxiv_id":"1805.10627","date":"2018-05-27","proceeding":"ACL 2018 7","authors":["Julia Kreutzer","Joshua Uyheng","Stefan Riezler"],"abstract":"We present a study on reinforcement learning (RL) from human bandit feedback\nfor sequence-to-sequence learning, exemplified by the task of bandit neural\nmachine translation (NMT). We investigate the reliability of human bandit\nfeedback, and analyze the influence of reliability on the learnability of a\nreward estimator, and the effect of the quality of reward estimates on the\noverall RL task. Our analysis of cardinal (5-point ratings) and ordinal\n(pairwise preferences) feedback shows that their intra- and inter-annotator\n$\\alpha$-agreement is comparable. Best reliability is obtained for standardized\ncardinal feedback, and cardinal feedback is also easiest to learn and\ngeneralize from. Finally, improvements of over 1 BLEU can be obtained by\nintegrating a regression-based reward estimator trained on cardinal feedback\nfor 800 translations into RL for NMT. This shows that RL is possible even from\nsmall amounts of fairly reliable human feedback, pointing to a great potential\nfor applications at larger scale.","url_abs":"http://arxiv.org/abs/1805.10627v3","url_pdf":"http://arxiv.org/pdf/1805.10627v3.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":"reliability-and-learnability-of-human-bandit","repo_url":"https://github.com/juliakreutzer/bandit-neuralmonkey","is_official":0,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":{"status":"ok","spdx":"BSD-3-Clause"}}],"tasks":[{"task_slug":"machine-translation","task_name":"Machine Translation"},{"task_slug":"nmt","task_name":"NMT"},{"task_slug":"reinforcement-learning","task_name":"Reinforcement Learning"},{"task_slug":"reinforcement-learning-1","task_name":"Reinforcement Learning (RL)"},{"task_slug":"translation","task_name":"Translation"},{"task_slug":"reinforcement-learning-2","task_name":"reinforcement-learning"}],"methods":[],"datasets_introduced":[{"slug":"humanmt","name":"HumanMT","full_name":""}],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1805.10627","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"1805.10627"}},"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/juliakreutzer/bandit-neuralmonkey","reach":{"status":"ok","spdx":"BSD-3-Clause"}}],"summary":{"unverified":7},"by_repo_kind":{"named_in_paper":{"samples":7,"ran":0,"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":"4702de42b8eb5525","entry":"evaluation","repo":"juliakreutzer/bandit-neuralmonkey","repo_kind":"named_in_paper","path":"neuralmonkey/learning_utils.py","file_url":"https://github.com/juliakreutzer/bandit-neuralmonkey/blob/HEAD/neuralmonkey/learning_utils.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"BSD-3-Clause","inline_ok":true,"mcp_get_code":{"code_sha256":"4702de42b8eb5525"}},{"code_sha256_prefix":"ee48d09163d7e0f8","entry":"get_resnet_by_type","repo":"juliakreutzer/bandit-neuralmonkey","repo_kind":"named_in_paper","path":"neuralmonkey/encoders/imagenet_encoder.py","file_url":"https://github.com/juliakreutzer/bandit-neuralmonkey/blob/HEAD/neuralmonkey/encoders/imagenet_encoder.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"BSD-3-Clause","inline_ok":true,"mcp_get_code":{"code_sha256":"ee48d09163d7e0f8"}},{"code_sha256_prefix":"e505daf00e73c61d","entry":"get_vgg_by_type","repo":"juliakreutzer/bandit-neuralmonkey","repo_kind":"named_in_paper","path":"neuralmonkey/encoders/imagenet_encoder.py","file_url":"https://github.com/juliakreutzer/bandit-neuralmonkey/blob/HEAD/neuralmonkey/encoders/imagenet_encoder.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"BSD-3-Clause","inline_ok":true,"mcp_get_code":{"code_sha256":"e505daf00e73c61d"}},{"code_sha256_prefix":"496fbab08f298f24","entry":"inverse_sigmoid_decay","repo":"juliakreutzer/bandit-neuralmonkey","repo_kind":"named_in_paper","path":"neuralmonkey/functions.py","file_url":"https://github.com/juliakreutzer/bandit-neuralmonkey/blob/HEAD/neuralmonkey/functions.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"BSD-3-Clause","inline_ok":true,"mcp_get_code":{"code_sha256":"496fbab08f298f24"}},{"code_sha256_prefix":"04265eb4f7053b66","entry":"noam_decay","repo":"juliakreutzer/bandit-neuralmonkey","repo_kind":"named_in_paper","path":"neuralmonkey/functions.py","file_url":"https://github.com/juliakreutzer/bandit-neuralmonkey/blob/HEAD/neuralmonkey/functions.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"BSD-3-Clause","inline_ok":true,"mcp_get_code":{"code_sha256":"04265eb4f7053b66"}},{"code_sha256_prefix":"75f62c423690e313","entry":"piecewise_function","repo":"juliakreutzer/bandit-neuralmonkey","repo_kind":"named_in_paper","path":"neuralmonkey/functions.py","file_url":"https://github.com/juliakreutzer/bandit-neuralmonkey/blob/HEAD/neuralmonkey/functions.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"BSD-3-Clause","inline_ok":true,"mcp_get_code":{"code_sha256":"75f62c423690e313"}},{"code_sha256_prefix":"e271b362c4222cd1","entry":"reinforce_score","repo":"juliakreutzer/bandit-neuralmonkey","repo_kind":"named_in_paper","path":"neuralmonkey/trainers/bandit_trainer.py","file_url":"https://github.com/juliakreutzer/bandit-neuralmonkey/blob/HEAD/neuralmonkey/trainers/bandit_trainer.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"BSD-3-Clause","inline_ok":true,"mcp_get_code":{"code_sha256":"e271b362c4222cd1"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}