Why Limit the Residual Stream to Layers and Not Tokens? Persistent Memory for Continuous Latent Reasoning
Quick Answer
The AGCLR model enhances the CoCoNuT paradigm by introducing a Gated Concept Stream, addressing the concept bottleneck in LLMs.
Quick Take
This innovation allows for persistent memory across reasoning passes, leading to improved performance on benchmarks like GSM8K and HotpotQA, with AGCLR outperforming vanilla CoCoNuT by resolving critical fact loss during reasoning. Code is available for further exploration.
Key Points
- AGCLR introduces a Gated Concept Stream to retain critical facts during reasoning.
- Performance improved on GSM8K and HotpotQA benchmarks compared to vanilla CoCoNuT.
- Concept bottleneck identified as a limitation in existing reasoning methods.
- Three learned gates (write, read, forget) control memory management in AGCLR.
- Code for AGCLR is publicly available for further research and application.
Paper Resources
Article Content
From source RSS / original summaryarXiv:2606. 07720v1 Announce Type: new Abstract: (LLMs) have demonstrated remarkable reasoning abilities on mathematical and multi-hop planning tasks. The CoCoNuT (Chain of Continuous Thought) paradigm~\cite{hao2024coconut} extends this by enabling models to reason in latent space, exploring multiple reasoning paths simultaneously rather than committing to a single chain early on. However, we identify a limitation we term the \textbf{concept bottleneck}.
At each reasoning pass, intermediate hidden states are overwritten, causing the model to lose critical facts computed in earlier steps as reasoning depth increases. We observe this empirically. On HotpotQA, vanilla CoCoNuT (10. 4\% EM) fails to improve over the CoT baseline (11. 0\% EM), and performance degrades with curriculum depth on GSM8K. To address this, we propose \textbf{AGCLR} (Adaptive Gated Continuous Latent Reasoning), which augments CoCoNuT with a \textit{Gated Concept Stream}.
A persistent residual memory maintained across all reasoning passes, controlled by three learned gates: a \textit{write} gate that commits intermediate facts to memory, a \textit{read} gate that retrieves relevant prior states, and a \textit{forget} gate that prunes irrelevant context. Evaluated on GSM8K, HotpotQA, and ProsQA using GPT-2 as our base model, AGCLR achieves consistent improvements across all types of datasets.
With the performance gap compounding as curriculum depth increases, directly resolving the concept bottleneck. Code available at https://anonymous. 4open. science/r/JJJJ/README. md
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →HOBA: Hierarchical On-Policy Bidding Agents for Adaptive Online Advertising
HOBA (Hierarchical On-policy Bidding Agents) is a novel hierarchical reinforcement learning framework that enhances online advertising bidding systems by improving adaptability and reducing hyperparameter tuning costs. It utilizes a for hyperparameter inference, a SARSA agent for expert model selection, and a dynamic expert pool for bid execution, achieving a +3.6% increase in target cost during large-scale deployment and outperforming state-of-the-art baselines on AuctionNet.