{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:7GCNP5QP72FF3U54YXHRRZORJ7","short_pith_number":"pith:7GCNP5QP","schema_version":"1.0","canonical_sha256":"f984d7f60ffe8a5dd3bcc5cf18e5d14ff0433282228ef2e479f7d5b5314e194e","source":{"kind":"arxiv","id":"2602.11202","version":3},"attestation_state":"computed","paper":{"title":"interwhen: A Generalizable Framework for Steering Reasoning Models with Test-time Verification","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Interwhen monitors reasoning traces in real time and steers models by verifying intermediate states against synthesized policy rules.","cross_cats":["cs.AI"],"primary_cat":"cs.LO","authors_text":"Amit Sharma, Ashmit Khandelwal, Maitreyi Swaroop, Nagarajan Natarajan, Prateek Chanda, Subbarao Kambhampati, Vijval Ekbote, Vineeth N. Balasubramanian, Vishak K Bhat","submitted_at":"2026-02-05T08:35:01Z","abstract_excerpt":"Reasoning models produce long traces of intermediate decisions and tool calls, making test-time verification important for ensuring correctness. Existing approaches either verify only the final answer, which misses early errors, or rely on branch-and-verify strategies that explore multiple trajectories. We introduce interwhen, a single-trajectory verification framework that steers model behavior by providing feedback on intermediate reasoning traces. It addresses two key challenges. First, given a set of verifiers, obtaining verifiable states from the reasoning trace typically requires prompt "},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2602.11202","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LO","submitted_at":"2026-02-05T08:35:01Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"75016f1e0a7d0c1baf16a53aac6d508421057089eae60d44d8b5c67c7dd4701e","abstract_canon_sha256":"9076f40e44ed1a7dcce7ae6da57701989e6497de5d209beca364a609cc3e529b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T02:45:05.313526Z","signature_b64":"0lWGh3uXDVxdGxXi38xYAXaSMNu0sUJ4phHf/q8NWSHNM606x474dZ7jSfn4JLdy34M2iIrVXr1vbREIp00HBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f984d7f60ffe8a5dd3bcc5cf18e5d14ff0433282228ef2e479f7d5b5314e194e","last_reissued_at":"2026-05-18T02:45:05.313031Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T02:45:05.313031Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"interwhen: A Generalizable Framework for Steering Reasoning Models with Test-time Verification","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Interwhen monitors reasoning traces in real time and steers models by verifying intermediate states against synthesized policy rules.","cross_cats":["cs.AI"],"primary_cat":"cs.LO","authors_text":"Amit Sharma, Ashmit Khandelwal, Maitreyi Swaroop, Nagarajan Natarajan, Prateek Chanda, Subbarao Kambhampati, Vijval Ekbote, Vineeth N. Balasubramanian, Vishak K Bhat","submitted_at":"2026-02-05T08:35:01Z","abstract_excerpt":"Reasoning models produce long traces of intermediate decisions and tool calls, making test-time verification important for ensuring correctness. Existing approaches either verify only the final answer, which misses early errors, or rely on branch-and-verify strategies that explore multiple trajectories. We introduce interwhen, a single-trajectory verification framework that steers model behavior by providing feedback on intermediate reasoning traces. It addresses two key challenges. First, given a set of verifiers, obtaining verifiable states from the reasoning trace typically requires prompt "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"On reasoning benchmarks where policies encode mathematical or logical constraints, interwhen achieves near-perfect accuracy for reasoning models using a fraction of the tokens of baselines. On agentic benchmarks with policy-based verifier generation, it enables improvements in task quality for SLMs without any finetuning, e.g., task completion rate of Qwen3-30B jumps from 32% to 87% on the telecom domain in tau2-bench.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the monitoring system can reliably poll and fork inference to recover accurate intermediate states from any reasoning trace, and that automatic synthesis from natural-language policies produces verifiers that are both correct and sufficiently complete to catch relevant violations.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"interwhen is a single-trajectory test-time verification system that polls reasoning traces, forks inference for intermediate states, synthesizes verifiers from policies including in Lean and z3, and steers models to near-perfect accuracy and higher task completion on benchmarks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Interwhen monitors reasoning traces in real time and steers models by verifying intermediate states against synthesized policy rules.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"c6615ab1160bba7b2b6714b4c1abd43cf05e0f0d2668ba5d02f5cf232b710976"},"source":{"id":"2602.11202","kind":"arxiv","version":3},"verdict":{"id":"29831cd7-5e9e-4728-b65e-53960f13bf65","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T07:24:50.951421Z","strongest_claim":"On reasoning benchmarks where policies encode mathematical or logical constraints, interwhen achieves near-perfect accuracy for reasoning models using a fraction of the tokens of baselines. On agentic benchmarks with policy-based verifier generation, it enables improvements in task quality for SLMs without any finetuning, e.g., task completion rate of Qwen3-30B jumps from 32% to 87% on the telecom domain in tau2-bench.","one_line_summary":"interwhen is a single-trajectory test-time verification system that polls reasoning traces, forks inference for intermediate states, synthesizes verifiers from policies including in Lean and z3, and steers models to near-perfect accuracy and higher task completion on benchmarks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the monitoring system can reliably poll and fork inference to recover accurate intermediate states from any reasoning trace, and that automatic synthesis from natural-language policies produces verifiers that are both correct and sufficiently complete to catch relevant violations.","pith_extraction_headline":"Interwhen monitors reasoning traces in real time and steers models by verifying intermediate states against synthesized policy rules."},"references":{"count":25,"sample":[{"doi":"","year":2025,"title":"Early stopping chain-of-thoughts in large language models.ArXiv, abs/2509.14004","work_id":"ec8d501d-19f8-40d2-967e-127c68240742","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"- A set of features (e.g., color, name, pet, book genre)","work_id":"e28c4709-c0e2-4236-b4e6-526d3c22911c","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Use any feedback to guide your reasoning until a complete solution is reached","work_id":"b132ada5-6526-4e66-9fc8-58349569a657","ref_index":6,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Do not stop responding until you’ve assigned each and every variable. # Final Answer Reporting Format ‘‘‘json { \"House 1\": { \"feature1\": \"value1\", \"feature2\": \"value2\", ... }, \"House 2\": { \"feature1\":","work_id":"911a9f3a-6ba2-449c-8bce-c5d6ab69ba59","ref_index":7,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"- A set of features (e.g., color, name, pet, book genre)","work_id":"75c56592-5c82-423a-8508-c55f06bec873","ref_index":8,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":25,"snapshot_sha256":"84a961bf08c26a5fd666b95fcb463559557691579308004e16e93d9fb29df750","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"92441b70d55915ae497f04ec9152a0af55205b9baaa80730a175e34594828e1d"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2602.11202","created_at":"2026-05-18T02:45:05.313111+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.11202v3","created_at":"2026-05-18T02:45:05.313111+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.11202","created_at":"2026-05-18T02:45:05.313111+00:00"},{"alias_kind":"pith_short_12","alias_value":"7GCNP5QP72FF","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"7GCNP5QP72FF3U54","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"7GCNP5QP","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7GCNP5QP72FF3U54YXHRRZORJ7","json":"https://pith.science/pith/7GCNP5QP72FF3U54YXHRRZORJ7.json","graph_json":"https://pith.science/api/pith-number/7GCNP5QP72FF3U54YXHRRZORJ7/graph.json","events_json":"https://pith.science/api/pith-number/7GCNP5QP72FF3U54YXHRRZORJ7/events.json","paper":"https://pith.science/paper/7GCNP5QP"},"agent_actions":{"view_html":"https://pith.science/pith/7GCNP5QP72FF3U54YXHRRZORJ7","download_json":"https://pith.science/pith/7GCNP5QP72FF3U54YXHRRZORJ7.json","view_paper":"https://pith.science/paper/7GCNP5QP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.11202&json=true","fetch_graph":"https://pith.science/api/pith-number/7GCNP5QP72FF3U54YXHRRZORJ7/graph.json","fetch_events":"https://pith.science/api/pith-number/7GCNP5QP72FF3U54YXHRRZORJ7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7GCNP5QP72FF3U54YXHRRZORJ7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7GCNP5QP72FF3U54YXHRRZORJ7/action/storage_attestation","attest_author":"https://pith.science/pith/7GCNP5QP72FF3U54YXHRRZORJ7/action/author_attestation","sign_citation":"https://pith.science/pith/7GCNP5QP72FF3U54YXHRRZORJ7/action/citation_signature","submit_replication":"https://pith.science/pith/7GCNP5QP72FF3U54YXHRRZORJ7/action/replication_record"}},"created_at":"2026-05-18T02:45:05.313111+00:00","updated_at":"2026-05-18T02:45:05.313111+00:00"}