SYNTHESIS NOTE
Topics›Reasoning by Reflection›this note

Can confidence patterns reveal overthinking versus underthinking?

This explores whether real-time confidence signals can diagnose when a reasoning model is trapped in redundant deliberation versus committing prematurely, and whether steering based on these signals can balance both failure modes.

Synthesis note · 2026-04-01 · sourced from Reasoning by Reflection

Overthinking and underthinking are dual failures, and existing methods that suppress one often induce the other. Suppressing reflective keywords or truncating reasoning length reduces overthinking but causes underthinking — the model doesn't explore enough. Forcing longer chains reduces underthinking but generates redundancy. ReBalance resolves this by treating confidence as a continuous diagnostic signal rather than using binary interventions.

The diagnostic: Confidence values correlate with reasoning behavior in interpretable ways:

The mechanism: From a small-scale dataset, identify reasoning steps indicating each mode. Aggregate their hidden states into reasoning mode prototypes. Compute a steering vector encoding the transition from overthinking to underthinking. A dynamic control function modulates the vector's strength and direction based on real-time confidence: pruning redundancy during overthinking, promoting exploration during underthinking.

Why it's training-free: The steering vector captures the model's inherent reasoning dynamics — it's extracted from the model's own hidden states, not trained. Because it operates on intrinsic representations, it generalizes across unseen data and tasks (math, QA, coding). This makes it plug-and-play across models from 0.5B to 32B.

Since Can we steer reasoning toward brevity without retraining?, ReBalance extends the activation-steering approach from length compression to reasoning quality management. ASC steers between verbose and concise modes; ReBalance steers between overthinking and underthinking — a qualitative distinction, not just quantitative.

Since Does more thinking time always improve reasoning accuracy?, ReBalance provides the dynamic mechanism the threshold finding calls for: instead of a fixed cutoff, confidence-based steering continuously adjusts the reasoning trajectory.

Inquiring lines that read this note 91

This note is a source for these research framings, grouped by the broader line of inquiry each explores. Scan the bold lines of inquiry; follow any specific question forward.

What factors determine AI persuasion effectiveness and what controls mitigate it? Does reasoning fine-tuning trade off other critical model abilities? How does transparency of agent reasoning affect effective oversight? Does chain-of-thought reasoning faithfully reflect actual model reasoning processes? Do reasoning traces accurately represent what models actually compute? Why does separating planning from execution enhance reasoning performance? How do different reward signals affect model reasoning and calibration differently? How can we determine if computational systems are genuinely conscious? When do additional thinking tokens stop improving reasoning performance? What determines whether visible reasoning traces improve task performance? How should conversational agents balance proactive goal-pursuit with user autonomy? How does offloading cognition to AI assistance affect learning retention? What makes reasoning models struggle with ill-posed and unfamiliar problems? When should retrieval be triggered to augment language model reasoning? Why does self-revision increase model confidence despite degraded accuracy? Does polished AI output mislead users about actual accuracy and expertise? Does preference-optimized training sacrifice accuracy and calibration for perceived quality? Does architectural recurrence enable computational capabilities that fixed-depth models cannot achieve? Do base models contain latent reasoning capabilities that training can activate? How can persona-based systems overcome their inherent limitations in modeling diversity? Why do autonomous agents fail and how well can they detect failures? When and why does aggregating multiple reasoning samples improve performance? Do language models genuinely understand or primarily match statistical patterns? What external mechanisms enable reliable self-improvement in language models? How can models learn to ask clarifying questions effectively? How do benchmark scores relate to real-world task completion? What actually drives chain-of-thought performance gains in reasoning tasks? How do transformer circuits and attention mechanisms enable or limit reasoning? Why do language models systematically fail at causal reasoning? How does scaling test-time compute compare to increasing model size? How does silent agreement emerge and undermine deliberation in multi-agent systems? Why do human-AI teams underperform compared to their individual components? Should agent evaluation measure execution quality beyond task success? How do clinicians evaluate the trustworthiness of AI medical recommendations? What training data structures and signals produce robust learning across distributions? Why doesn't LLM knowledge translate to improved reasoning outputs?

Related concepts in this collection 4

This note in its neighbourhood — explore the map, then jump to a related concept in the list below.

Concept map
12 direct connections · 105 in 2-hop network ·medium cluster Open in graph ↗
Can confidence patterns reveal overthinking vers… Can we steer reasoning toward brevity without retr… Does more thinking time always improve reasoning a… Do reasoning models switch between ideas too frequ… When should an agent actually stop and deliberate?

Click a node to walk · click center to open · click Open in graph to see this note in the full knowledge graph

your link semantically near linked from elsewhere

Related papers in this collection 8

Papers most semantically related to this note, ranked by cosine similarity in the embedding space.

Search by related questions 4

Suggested questions this note speaks to — click to search the collection, or type your own.

How do verbose and concise reasoning occupy different regions in activation space?What is the critical thinking token threshold beyond which accuracy degrades?Can penalizing reasoning transitions fix underthinking without fine-tuning models?When should action deliberation trigger during reasoning steps?
Original note title

ReBalance uses confidence as continuous indicator to dynamically steer between overthinking and underthinking — training-free balanced reasoning via hidden state steering vectors