REVIEW 5 cited by
Deliberation in Latent Space via Differentiable Cache Augmentation
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
Techniques enabling large language models (LLMs) to "think more" by generating and attending to intermediate reasoning steps have shown promise in solving complex problems. However, the standard approaches generate sequences of discrete tokens immediately before responding, and so they can incur significant latency costs and be challenging to optimize. In this work, we demonstrate that a frozen LLM can be augmented with an offline coprocessor that operates on the model's key-value (kv) cache. This coprocessor augments the cache with a set of latent embeddings designed to improve the fidelity of subsequent decoding. We train this coprocessor using the language modeling loss from the decoder on standard pretraining data, while keeping the decoder itself frozen. This approach enables the model to learn, in an end-to-end differentiable fashion, how to distill additional computation into its kv-cache. Because the decoder remains unchanged, the coprocessor can operate offline and asynchronously, and the language model can function normally if the coprocessor is unavailable or if a given cache is deemed not to require extra computation. We show experimentally that when a cache is augmented, the decoder achieves lower perplexity on numerous subsequent tokens. Furthermore, even without any task-specific training, our experiments demonstrate that cache augmentation consistently reduces perplexity and improves performance across a range of reasoning-intensive tasks.
Forward citations
Cited by 5 Pith papers
-
SkillSmith: Learning to Compose Parametric Skills and Textual Knowledge
SkillSmith, an LLM augmented to ingest prefix-weights and text, generates target-task prefix-weights that beat text-only and weight-only baselines, especially as fine-tuning initialization.
-
The Vision Wormhole: Latent-Space Communication in Heterogeneous Multi-Agent Systems
Reasoning messages between heterogeneous VLMs can be routed through the image-token span: a distilled universal codec plus affine alignment transmits latent traces across model families, cutting wall-clock time in sma...
-
LatentChem: From Textual CoT to Latent Thinking in Chemical Reasoning
LatentChem reasons in continuous latent space for chemistry, achieving a 59.88% non-tie win rate over explicit CoT on ChemCoTBench with a 10.84x average reduction in reasoning overhead.
-
Latent Collaboration in Multi-Agent Systems
Replacing text inter-agent dialogue with direct transfer of hidden-state (KV-cache) representations cuts output tokens by ~70-84%, speeds inference ~4x, and keeps multi-agent accuracy roughly on par or slightly better.
-
Token Assorted: Mixing Latent and Text Tokens for Improved Language Model Reasoning
Partially replacing chain-of-thought tokens with discrete latent tokens during fine-tuning improves LLM reasoning accuracy and reduces generated token count.
Discussion (0). Continue with ORCID to comment.