--- title: 'RecLoop: Closed-Loop Recommender Simulation' url: https://www.emergentmind.com/topics/recloop type: topic --- # RecLoop: Closed-Loop Recommender Simulation Searching arXiv for the RecLoop paper and closely related recommendation-feedback-loop work. RecLoop is a closed-loop simulation framework for analyzing how recommender systems and user feedback co-evolve over repeated interaction cycles, with a specific focus on whether generative recommenders deepen information cocoons [2606.17707]. In the framework, recommendation is not treated as a one-shot offline prediction problem: a recommender produces exposure lists, LLM-powered user agents select from those lists, the resulting interactions are appended to user histories, and the recommender is retrained on the updated data. This repeated feedback protocol is used to compare traditional sequential recommenders and generative recommenders under identical long-term conditions, and to measure both conventional exposure-level concentration and a model-internal phenomenon termed Code-Space Structural Cocoon [2606.17707]. ## 1. Problem setting and conceptual scope RecLoop is designed around the claim that static offline evaluation cannot capture feedback-loop effects in recommender systems, because recommendations affect what users click and those interactions then retrain the model [2606.17707]. Its central research question is whether generative recommenders, which replace atomic item IDs with Semantic ID sequences, alter the dynamics of information cocoon formation relative to traditional sequential recommenders [2606.17707]. The framework studies two recommendation paradigms. The first is the traditional sequential paradigm, represented by SASRec and Mamba4Rec, which use atomic item IDs and embedding-based next-item scoring. The second is the generative paradigm, represented by TIGER and OneRec, in which recommendation proceeds through generated discrete code sequences rather than direct scoring over atomic IDs [2606.17707]. This distinction matters because the paper argues that cocooning in generative systems may emerge not only in exposed items, but also inside the generated code space itself [2606.17707]. A key conceptual contribution is therefore the shift from observing only visible recommendation lists to also analyzing internal representational concentration. This suggests that recommender feedback loops should be treated as dynamics over both user exposure and model-internal generation structure, rather than as a purely ranking-level phenomenon [2606.17707]. ## 2. Closed-loop simulation dynamics For each user \(u\), the simulation begins from an initial history \(\mathcal{S}_u^0\). At cycle \(t\), the recommender \(f_\theta^t\) consumes the accumulated history and outputs a top-\(K\) exposure list, \[ \mathcal{E}_u^t = f_\theta^t(\mathcal{S}_u^t). \] A user agent then selects one item \(i_u^t\), after which the interaction history is updated as \[ \mathcal{S}_u^{t+1} = \mathcal{S}_u^t \oplus i_u^t. \] After all users respond in a cycle, the recommender is retrained on the updated data, and the procedure repeats for 15 feedback cycles [2606.17707]. This formulation makes cocoon formation an emergent property of repeated recommendation-training coupling rather than a static property of a trained model. The protocol uses a top-\(K=5\) exposure list at each cycle, deterministic user feedback with temperature \(=0\), and retraining after every cycle [2606.17707]. The closed-loop design is intended to emulate the endogenous reinforcement process often absent from offline benchmarks. A plausible implication is that RecLoop functions as a controlled testbed for separating short-term recommendation quality from long-term distributional effects, although the paper frames this point through the repeated recommendation-selection-retraining protocol rather than as a general formal theory [2606.17707]. ## 3. LLM-powered user simulators Each simulated user is represented by an LLM agent with four components: a dynamic user profile, dual memory, periodic reflection, and an action or decision module [2606.17707]. The base profile \(P_u^{\mathrm{base}}\) is constructed offline from the user’s pre-simulation history by prompting an LLM to write a first-person consumer persona. The stage-wise profile is initialized by \[ P_u^0 = P_u^{\mathrm{base}}, \] and is refreshed every \(\Delta\) cycles through reflection [2606.17707]. The memory system is split into short-term and long-term components. Short-term memory \(M_u^t\) is a sliding window over the most recent \(W\) cycles, while long-term memory is the full interaction history \(\mathcal{S}_u^t\). In the reported experiments, \(W=5\) and the reflection interval is \(\Delta=5\) [2606.17707]. Reflection summarizes the accumulated interaction history into an updated profile and is intended to capture evolving interests without overreacting to an isolated action [2606.17707]. The action module receives a structured prompt consisting of the stage-wise profile, recent memory, the current exposure list, and a constraint layer listing the exact valid item IDs. Its structured output is \[ (i_u^t, r_u^t) = \mathrm{LLM}\big(\pi(P_u^t, M_u^t, \mathcal{E}_u^t)\big), \] where \(i_u^t\) is the selected item and \(r_u^t\) is the reason [2606.17707]. The constraint layer is specifically used to ensure that the LLM chooses only from exposed candidates and does not hallucinate item IDs [2606.17707]. In the simulation protocol, the user simulator is Qwen3-8B [2606.17707]. This places RecLoop in the class of agent-based evaluation frameworks, but here the agent architecture is tightly coupled to recommendation feedback and retraining dynamics rather than being used only for standalone behavioral emulation. ## 4. Experimental configuration The experiments use two Amazon review datasets with hierarchical categories and 5-core preprocessing under chronological ordering of user histories [2606.17707]. | Dataset | Users / Items / Interactions | Additional details | |---|---:|---| | Office Products | 4,905 / 2,420 / 53,258 | Avg. length: 10.86; Categories: \((5/11/44/152)\) | | Toys and Games | 19,412 / 11,924 / 167,597 | Avg. length: 8.63; Categories: \((11/41/191)\) | Four recommender models are evaluated under a common simulation protocol. The traditional sequential baselines are SASRec and Mamba4Rec. The generative recommenders are TIGER and OneRec, with OneRec evaluated at scales 0.5B, 1.5B, and 3B. For comparability, the implementation uses the MiniOneRec codebase with supervised fine-tuning only [2606.17707]. For generative recommenders, the study compares two tokenization strategies. Semantic IDs (SID) are obtained by training RQ-VAE on text embeddings from titles and descriptions, whereas Collaborative IDs (CID) are obtained by training RQ-VAE on SASRec item embeddings. The architecture, code depth, codebook size, and hyperparameters are held fixed; only the input representation differs [2606.17707]. The simulation uses 15 cycles, top-\(K=5\) exposures, user history initialization capped at length 50, short-term memory window \(W=5\), reflection interval \(\Delta=5\), and NVIDIA A800 80GB hardware [2606.17707]. These design choices ensure that paradigm differences are studied under a common feedback regime. ## 5. Metrics and the Code-Space Structural Cocoon RecLoop evaluates cocoon formation with four standard exposure-level metrics and one new code-space metric family [2606.17707]. The exposure-level metrics are intra-user diversity, inter-user homogenization, system-level coverage, and exposure concentration. Intra-user diversity is measured by category entropy. For user \(u\), cycle \(t\), and category level \(l\), \[ E_{t,l} = -\frac{1}{|U|}\sum_{u \in U}\sum_{c \in \mathcal{C}_l} p_{t,u}(c)\log p_{t,u}(c), \] where \(p_{t,u}(c)\) is the fraction of items in user \(u\)’s exposure list belonging to category \(c\). Lower entropy indicates narrower exposure [2606.17707]. Inter-user homogenization is measured by average pairwise Jaccard similarity, \[ \bar{J}^t = \frac{1}{\binom{|\mathcal{U}|}{2}}\sum_{u