Paper deep dive
Towards Reversible Forgetting: Managing Obsolete Knowledge in Continual Enterprise AI Agents
Nilutpaul Sarker Yash, Tirtho Roy, Ushashi Bhattacharjee
Intelligence
Status: succeeded | Model: Gemma-4-26B-A4B | Prompt: intel-v1 | Confidence: 93%
Last extracted: 8/20/2026, 3:43:14 AM
Summary
The paper proposes 'reversible forgetting' as a framework for managing obsolete knowledge in continual enterprise AI agents. It introduces the Hysteretic Reversible Memory Controller (HRMC), which manages memory states (active, dormant, retired) using asymmetric thresholds and hysteresis to prevent state oscillation. The framework allows for the suppression of obsolete knowledge without permanent erasure, enabling reactivation when relevance returns, illustrated with a finance domain example.
Entities (10)
Relation Signals (9)
Hysteretic Reversible Memory Controller → manages → Active State
confidence 95% · three operational memory states: active, dormant, and retired
Hysteretic Reversible Memory Controller → manages → Dormant State
confidence 95% · three operational memory states: active, dormant, and retired
Hysteretic Reversible Memory Controller → manages → Retired State
confidence 95% · three operational memory states: active, dormant, and retired
Reversible Forgetting → proposes → Hysteretic Reversible Memory Controller
confidence 95% · We instantiate the framework as a Hysteretic Reversible Memory Controller
Finance → illustrates → Reversible Forgetting
confidence 93% · Finance illustrates the idea: knowledge useful under one market regime may become harmful under another
Catastrophic Forgetting → iscontrastedwith → Reversible Forgetting
confidence 92% · Continual learning has traditionally treated forgetting as a failure... We argue that this objective is incomplete
Obsolete Knowledge → causes → Negative Transfer
confidence 90% · Indiscriminate retention can allow obsolete knowledge to influence decisions, creating negative transfer
Hysteretic Reversible Memory Controller → uses →
Cypher Suggestions (0)
No Cypher suggestions yet.
Abstract
Abstract:Continual learning has traditionally treated forgetting as a failure, emphasizing preservation of previously acquired knowledge as environments evolve. We argue that this objective is incomplete for enterprise AI agents operating in non-stationary environments, where customers, policies, tools, workflows, regulations, and market conditions change over time. Indiscriminate retention can allow obsolete knowledge to influence decisions, creating negative transfer and operational risk. We therefore propose reversible forgetting: a conceptual framework with three operational memory states: active, dormant, and retired, and a reactivation transition that can restore dormant knowledge when its relevance returns. We instantiate the framework as a Hysteretic Reversible Memory Controller that accumulates relevance evidence, uses asymmetric thresholds to prevent state oscillation, tests reactivation in shadow mode, and gates retirement through policy. The framework reduces the influence of obsolete information without conflating temporary suppression with permanent erasure. Finance illustrates the idea: knowledge useful under one market regime may become harmful under another yet regain relevance when similar conditions recur.
Tags
Links
- Source: https://arxiv.org/abs/2608.18177v1
- Canonical: https://arxiv.org/abs/2608.18177v1
Trouble viewing inline? Open PDF directly →
Full Text
17,272 characters extracted from source content.
Expand or collapse full text
Towards Reversible Forgetting: Managing Obsolete Knowledge in Continual Enterprise AI Agents Nilutpaul Sarker Yash Tirtho Roy Ushashi Bhattacharjee Abstract Continual learning has traditionally treated forgetting as a failure, emphasizing preservation of previously acquired knowledge as environments evolve. We argue that this objective is incomplete for enterprise AI agents operating in non-stationary environments, where customers, policies, tools, workflows, regulations, and market conditions change over time. Indiscriminate retention can allow obsolete knowledge to influence decisions, creating negative transfer and operational risk. We therefore propose reversible forgetting: a conceptual framework with three operational memory states—active, dormant, and retired—and a reactivation transition that can restore dormant knowledge when its relevance returns. We instantiate the framework as a Hysteretic Reversible Memory Controller that accumulates relevance evidence, uses asymmetric thresholds to prevent state oscillation, tests reactivation in shadow mode, and gates retirement through policy. The framework reduces the influence of obsolete information without conflating temporary suppression with permanent erasure. Finance illustrates the idea: knowledge useful under one market regime may become harmful under another yet regain relevance when similar conditions recur. Keywords: continual learning; enterprise AI agents; reversible forgetting; agent memory; knowledge lifecycle; distribution shift; AI governance. TL;DR: Enterprise agents should reversibly suppress obsolete knowledge rather than retain or erase it indiscriminately; HRMC operationalizes this position through auditable memory states, hysteresis, shadow-tested reactivation, and policy-gated retirement. 1 Introduction Continual learning asks how a system can acquire knowledge without losing useful prior capabilities (11; 6; 16). For enterprise agents, however, older knowledge should not always remain operational: customers, policies, tools, workflows, regulations, and markets change. This creates a second failure mode beyond catastrophic forgetting: obsolete workflows, strategies, policies, or experiences can cause negative transfer when retrieval continues to expose them. Lifelong-agent research identifies memory evolution as an open problem (20), and insight-governance work shows that stale verbal experience can harm later decisions (5). Position. Continual enterprise agents should support reversible forgetting: obsolete or contextually irrelevant knowledge should be suppressible from normal use without necessarily being erased, while preserving provenance and later reactivation. This reframes continual learning as management of a knowledge lifecycle rather than maximization of retention. Our argument makes four contributions: (1) we distinguish catastrophic forgetting from intentional, beneficial suppression; (2) we introduce active, dormant, and retired memory states connected by auditable suppression and reactivation transitions; (3) we propose a concrete controller with temporal evidence accumulation, hysteresis, shadow reactivation, and policy-gated retirement; and (4) we outline a research and benchmarking agenda for detection, granularity, retention, reactivation, evaluation, safety, and governance. 2 From catastrophic forgetting to beneficial forgetting Continual learning studies sequential adaptation while maintaining useful earlier capabilities (18; 15). Its central failure is catastrophic forgetting: new learning damages knowledge that remains valuable. Enterprise agents face the converse risk. A redesigned system, deprecated tool, superseded rule, or changed operating regime can make preserved knowledge misleading. Thus, accidental loss of useful knowledge is catastrophic, but controlled suppression of obsolete knowledge can be adaptive. The question becomes what should influence the agent now, remain dormant, or be retired. 2.1 Relationship to prior and adjacent work Catastrophic-forgetting methods preserve useful capabilities (8), while backward transfer improves old tasks using new knowledge (9); our concern is whether old knowledge should influence the current task at all. Machine unlearning removes designated training influence (3); reversible forgetting retains recoverability and cannot satisfy required erasure. Concept-drift methods detect distributional change (7; 17), whereas agent-memory and experience systems govern storage, retrieval, and transfer (19; 14). The closest precedent is the curation loop of 5, which validates and governs verbal insights and highlights stale experience in finance. Knowledge activation packages institutional knowledge into reusable units (2), and organizational reviews emphasize governance (1). HRMC instead makes a specific, falsifiable proposal: active–dormant–retired states, asymmetric thresholds, persistence counters, shadow-tested reactivation, policy-gated retirement, and a transition ledger. We claim these as design hypotheses, not demonstrated advantages. 3 Why enterprise agents need reversible forgetting Enterprise agents accumulate facts, interactions, tools, workflows, and policies (12; 20). The usual update Mt+1=Mt∪ΔMtM_t+1=M_t∪ M_t mixes memories acquired under incompatible assumptions. This is consequential because organizational knowledge is authoritative, versioned, access-controlled, and tied to owners (1; 2). Yet audits, incident reconstruction, and recurring conditions require historical versions. Reversible forgetting targets this combination of suppression, recoverability, provenance, and governance. 4 Reversible forgetting framework We treat enterprise memory as a lifecycle rather than a retention/deletion binary. A shift detector produces context RtR_t; a controller then assesses which memories should remain active, become dormant, or move toward retirement (Figure 1). Suppression is not deletion. Figure 1: Reversible-forgetting lifecycle: change triggers relevance assessment; dormant knowledge can return to active use, whereas retirement requires a policy gate. Memory states. Active knowledge supports normal retrieval; dormant knowledge is suppressed but recoverable; reactivation returns it to active use; and retired knowledge cannot be autonomously restored. A provenance stub may remain subject to policy. Retirement is not erasure: legally mandated deletion requires a separate unlearning or records-management process. The objective is to ensure that the right knowledge influences the agent at the right time. Formalizing relevance. For a memory mim_i at time t, consider the conceptual score Rt(mi)=αSt(mi)+βUt(mi)+γC(mi)−δAt(mi)−λHt(mi),R_t(m_i)=α S_t(m_i)+β U_t(m_i)+γ C(m_i)-δ A_t(m_i)-λ H_t(m_i), (1) where S, U, C, A, and H denote contextual similarity, utility, reliability, staleness, and harm. Implementations might estimate them from embeddings, counterfactual performance, source history, superseding events, and policy violations. The score illustrates candidate signals rather than a validated estimator; proxies and thresholds must be calibrated by memory layer and policy domain. 4.1 Hysteretic Reversible Memory Controller A direct threshold rule can repeatedly move a memory between states. The Hysteretic Reversible Memory Controller (HRMC) instead maintains R¯t(mi)=ρR¯t−1(mi)+(1−ρ)Rt(mi), R_t(m_i)=ρ R_t-1(m_i)+(1-ρ)R_t(m_i), (2) with τ↓<τ↑ _ < _ . Dormancy requires L↓L_ consecutive low scores; reactivation requires L↑L_ high scores and shadow gain above ϵε. Hysteresis and persistence limit state “flapping.” After TretireT_retire, retirement still requires policy or owner approval. Every transition and its evidence enter a ledger. Algorithm 1 leaves estimators and policies domain-specific. Algorithm 1 Hysteretic Reversible Memory Controller (HRMC) 1: Memory mim_i, context RtR_t, state ziz_i, score R¯t−1 R_t-1, counters ci−c_i^-, ci+c_i^+ 2: ri←Relevance(mi,Rt)r_i← Relevance(m_i,R_t) 3: R¯t←ρR¯t−1+(1−ρ)ri R_t←ρ R_t-1+(1-ρ)r_i 4: if zi=Activez_i= Active then 5: ci−←UpdateLowCount(R¯t<τ↓)c_i^-← UpdateLowCount( R_t< _ ) 6: if ci−≥L↓c_i^-≥ L_ or PolicyBlock(mi,Rt) PolicyBlock(m_i,R_t) then 7: Transition(mi,Dormant,evidence) Transition(m_i, Dormant,evidence) 8: end if 9: else if zi=Dormantz_i= Dormant then 10: ci+←UpdateHighCount(R¯t>τ↑)c_i^+← UpdateHighCount( R_t> _ ) 11: if ci+≥L↑c_i^+≥ L_ and ShadowGain(mi,Rt)>ϵ ShadowGain(m_i,R_t)>ε then 12: Transition(mi,Active,evidence) Transition(m_i, Active,evidence) 13: else if DormantAge(mi)>Tretire DormantAge(m_i)>T_retire and ApproveRetirement(mi) ApproveRetirement(m_i) then 14: Transition(mi,Retired,evidence) Transition(m_i, Retired,evidence) 15: end if 16: end if 17: AppendLedger(mi,zi,R¯t,evidence) AppendLedger(m_i,z_i, R_t,evidence) HRMC aims to preserve current-context utility while reducing harmful retrieval, transitions, and storage subject to governance. Similarity may nominate a memory for return, but counterfactual utility must justify reactivation. Granularity. Suppression must match the memory layer: retrieval filtering for episodes, versioning for facts and policies, preference decay for tools, retirement for workflows, and parameter adaptation for model behavior. The system must decide both whether knowledge is relevant and how to reduce its influence. 5 Finance as an illustrative enterprise domain Finance makes the idea concrete; stale verbal insights already show costs in this domain (5). A risk agent may learn thresholds and workflows M1M_1 in a low-volatility regime R1R_1. Under crisis regime R2R_2, these relationships can create excessive alerts. Rather than retain or delete M1M_1, the controller makes it dormant while M2M_2 becomes active; if R3≈R1R_3≈ R_1, it reactivates M1M_1 (Figure 2). Figure 2: Recurring finance regimes: M1M_1 is active in R1R_1, becomes dormant while M2M_2 serves R2R_2, and is reactivated when R3R_3 resembles R1R_1. 6 Governance and failure modes Enterprise systems must explain why knowledge stopped influencing behavior. A forgetting ledger records (mi,stateold,statenew,reason,time,evidence).(m_i,\,state_old,\,state_new,\,reason,\,time,\,evidence). This supports reconstruction, impact analysis, and restoration. Risks include premature suppression, false shift detection, unbounded dormant storage, failed reactivation, adversarial feedback, and compliance conflict. Dormancy cannot satisfy erasure because content remains recoverable. Access control, retention schedules, legal holds, deletion duties, and human override must constrain transitions, aligning the controller with lifecycle risk governance (13). 7 Research and benchmarking agenda A practical theory raises six questions: detection of obsolescence; granularity across examples, facts, tools, workflows, policies, and parameters; retention limits for dormant knowledge; reactivation evidence; evaluation separating beneficial suppression from catastrophic forgetting; and governance over transition authority. Benchmarks need recurring environments. Retention and transfer vary across sequential embodied tasks (4; 10); the sequence E1→E2→E3→E1E_1\!→\!E_2\!→\!E_3\!→\!E_1 additionally tests whether suppression helps during change and whether prior utility returns upon recurrence. A benchmark should report current-context utility, harmful-retention cost relative to an oracle policy, reactivation recovery, and transition cost. Governance measures should cover unsupported suppression, policy violations, audit completeness, and actual deletion. HRMC enables ablations of hysteresis, evidence persistence, shadow evaluation, and policy gating. Retain-all, fixed-window, least-recently-used, permanent-delete, and oracle-context baselines make the position falsifiable: does governed reversible control improve the utility–interference trade-off? References Anjelia et al. (2025) S. R. Anjelia, D. I. Sensuse, and S. Lusa AI agents for organizational knowledge retrieval and sharing: a systematic literature review. International Journal of Advances in Data and Information Systems. External Links: Document Cited by: §2.1, §3. Bakal (2026) G. Bakal Knowledge activation: AI skills as the institutional knowledge primitive for agentic software development. arXiv preprint arXiv:2603.14805. External Links: Document Cited by: §2.1, §3. Bourtoule et al. (2021) L. Bourtoule, V. Chandrasekaran, C. A. Choquette-Choo, H. Jia, A. Travers, B. Zhang, D. Lie, and N. Papernot Machine unlearning. In 2021 IEEE Symposium on Security and Privacy, p. 141–159. External Links: Document Cited by: §2.1. Cudrano et al. (2024) P. Cudrano, X. Luo, and M. Matteucci The empirical impact of forgetting and transfer in continual visual odometry. arXiv preprint arXiv:2406.01797. External Links: Document Cited by: §7. Cui et al. (2026) Y. Cui, X. Zhang, Y. Zhang, L. Shao, X. Shi, G. Wang, and P. He Closing the feedback loop: from experience extraction to insight governance in verbal reinforcement learning. arXiv preprint arXiv:2606.17591. External Links: Document Cited by: §1, §2.1, §5. De Lange et al. (2022) M. De Lange, R. Aljundi, M. Masana, S. Parisot, X. Jia, A. Leonardis, G. Slabaugh, and T. Tuytelaars A continual learning survey: defying forgetting in classification tasks. IEEE Transactions on Pattern Analysis and Machine Intelligence 44 (7), p. 3366–3385. External Links: Document Cited by: §1. Gama et al. (2014) J. Gama, I. Zliobaite, A. Bifet, M. Pechenizkiy, and A. Bouchachia A survey on concept drift adaptation. ACM Computing Surveys 46 (4), p. 44:1–44:37. External Links: Document Cited by: §2.1. Kirkpatrick et al. (2017) J. Kirkpatrick, R. Pascanu, N. Rabinowitz, J. Veness, G. Desjardins, A. A. Rusu, K. Milan, J. Quan, T. Ramalho, A. Grabska-Barwinska, et al. Overcoming catastrophic forgetting in neural networks. Proceedings of the National Academy of Sciences 114 (13), p. 3521–3526. External Links: Document Cited by: §2.1. Lin et al. (2022) S. Lin, L. Yang, D. Fan, and J. Zhang Beyond not-forgetting: continual learning with backward knowledge transfer. arXiv preprint arXiv:2211.00789. External Links: Document Cited by: §2.1. Meng et al. (2025) Y. Meng, Z. Bing, X. Yao, K. Chen, K. Huang, Y. Gao, F. Sun, and A. Knoll Preserving and combining knowledge in robotic lifelong reinforcement learning. Nature Machine Intelligence. External Links: Document Cited by: §7. Parisi et al. (2019) G. I. Parisi, R. Kemker, J. L. Part, C. Kanan, and S. Wermter Continual lifelong learning with neural networks: a review. Neural Networks 113, p. 54–71. External Links: Document Cited by: §1. Sapkota et al. (2025) R. Sapkota, K. I. Roumeliotis, and M. Karkee AI agents vs. agentic AI: a conceptual taxonomy, applications and challenges. Information Fusion. External Links: Document Cited by: §3. Tabassi (2023) E. Tabassi Artificial intelligence risk management framework (AI RMF 1.0). Technical report Technical Report NIST AI 100-1, National Institute of Standards and Technology. External Links: Document Cited by: §6. Tang et al. (2025) X. Tang, T. Qin, T. Peng, Z. Zhou, Y. Shao, T. Du, X. Wei, P. Xia, F. Wu, H. Zhu, et al. Agent KB: leveraging cross-domain experience for agentic problem solving. arXiv preprint arXiv:2507.06229. External Links: Document Cited by: §2.1. van de Ven et al. (2024) G. M. van de Ven, N. Soures, and D. Kudithipudi Continual learning and catastrophic forgetting. Note: Book chapter External Links: Document Cited by: §2. Wang et al. (2024) L. Wang, X. Zhang, H. Su, and J. Zhu A comprehensive survey of continual learning: theory, method and application. IEEE Transactions on Pattern Analysis and Machine Intelligence. External Links: Document Cited by: §1. Weng (2024) W. Weng Handling non-stationary data streams under complex environments. Ph.D. Thesis, Nanyang Technological University. External Links: Document Cited by: §2.1. Wickramasinghe et al. (2024) B. Wickramasinghe, G. Saha, and K. Roy Continual learning: a review of techniques, challenges, and future directions. IEEE Transactions on Artificial Intelligence. External Links: Document Cited by: §2. Xiong et al. (2025) Z. Xiong, Y. Lin, W. Xie, P. He, J. Tang, H. Lakkaraju, and Z. Xiang How memory management impacts llm agents: an empirical study of experience-following behavior. arXiv preprint arXiv:2505.16067. Cited by: §2.1. Zheng et al. (2025) J. Zheng, C. Shi, X. Cai, Q. Li, D. Zhang, C. Li, D. Yu, and Q. Ma Lifelong learning of large language model based agents: a roadmap. IEEE Transactions on Pattern Analysis and Machine Intelligence. External Links: Document Cited by: §1, §3.